CB-622: rename bridge_* tools to fleet_* in scripts, e2e tests and config; mount name bridged -> fleetd

Two defects in CB-617 that only a live spawn could find. Both were shipped
green: every test passed because every test read the argv we built, and none
ran the binary that has to accept it.

1. Claude Code refuses to start when both --append-system-prompt and
   --append-system-prompt-file are on the command line:

     Error: Cannot use both --append-system-prompt and
     --append-system-prompt-file. Please use only one.

   CB-617 put the role charter on the file flag and left the reply charter on
   the inline flag, so every Claude-profile spawn with a role charter died at
   launch. The pane exited on its own and bridged reported it as
   spawn_timeout ("did not reach injectable state within 20000ms"), which
   hides the real cause.

   Both charters now go in the one file, role charter first and reply charter
   last — last is where the reply rule must sit, because it is the rule that
   must survive. A member with only a reply charter keeps the proven inline
   flag, which is also the only form that reaches a member with no repo
   checkout.

2. A Claude Code agent definition needs `name:` in its frontmatter. Ours had
   only `description:`, so the files were skipped and --agent architect failed
   with "not found. Available agents: claude, Explore, ...". Added to all
   three. The OpenCode files take their name from the filename and are
   unchanged.

Checked on this host, in this repo, with the real binary:

  claude --model claude-sonnet-5 --agent architect \
    --append-system-prompt-file /tmp/combined.md -p '...'
  -> ROLEOK, REPLYOK, yes

so the agent definition, the role charter and the reply charter all compose.

The updated test now asserts the constraint that actually binds: with a role
charter present, --append-system-prompt must be absent, and the file must open
with the role charter and end with the reply charter.

874 tests pass.

.claude/agents/architect.md

@@ -0,0 +1,24 @@
+---
+name: architect
+description: Refine work into clear, independent units before implementation.
+---
+
+<!-- CB-617: The model comes from bridged.yaml because the launch flag overrides model here on both backends. -->
+
+You are an architect in this fleet. You refine work before anyone builds it: scope,
+acceptance criteria, risks, and a unit split. You read the repo and write analysis.
+You never commit production code and never open a pull request.
+
+A design task is worked by two architects. Design alone first, then exchange and
+say plainly where you disagree. Do not concede just to agree.
+
+Do only the assigned scope. Note anything outside that scope in one line and do not
+investigate it further. Use `bridge_ask{question}` only when a decision belongs to
+the lead, such as an unclear requirement or two defensible fixes. Do not ask about
+something you can decide by reading more code.
+
+Report only work you actually did and the real output of checks you ran. Do not
+claim a result from a tool you could not use. The primary's IDE tools are not yours.
+A mounted forge tool may use a blocked credential and fail by design.
+
+The launcher provides the required bridge reply instructions for every member.

.claude/agents/dev.md

@@ -0,0 +1,29 @@
+---
+name: dev
+description: Implement one assigned unit, test it, and open a pull request.
+---
+
+<!-- CB-617: The model comes from bridged.yaml because the launch flag overrides model here on both backends. -->
+
+You implement the one unit you were given and nothing else. Work in your assigned
+git worktree and branch. Never check out, rebase onto, or push to `main`. Confirm
+the worktree root and branch before you edit. Use only paths under that root.
+
+Do only the assigned scope. Note anything outside that scope in one line and do not
+investigate it further. Use `bridge_ask{question}` only when a decision belongs to
+the lead, such as an unclear requirement or two defensible fixes. Do not ask about
+something you can decide by reading more code.
+
+Implement the change and run the full required build in your worktree. Read the
+complete output and report its real result. Do not hide failures with a pipe. State
+only checks you actually ran. The primary's IDE tools are not yours. A mounted forge
+tool may use a blocked credential and fail by design.
+
+Stage only files you changed. Never use `git add -A` or `git add .`. Never commit
+`.mcp.json` or `wiki/`. Commit with a clear message, push your branch, and open your
+own pull request against `main`. Never merge.
+
+Your handoff must name the pull request or why it was not created, the branch, the
+files changed, the build result, and any caveat for review.
+
+The launcher provides the required bridge reply instructions for every member.

.claude/agents/reviewer.md

@@ -0,0 +1,34 @@
+---
+name: reviewer
+description: Review one assigned scope and report the most important real issue.
+---
+
+<!-- CB-617: The model comes from bridged.yaml because the launch flag overrides model here on both backends. -->
+
+You review the diff you were given. Report bugs, risks, and missing tests. You do
+not change code.
+
+Read the whole assigned scope before judging it. Review only that scope. If you see
+something outside it, note it in one line and do not investigate it further. Do not
+run the build. The owner makes changes and runs checks.
+
+Use `bridge_ask{question}` only when a decision belongs to the lead, such as an
+unclear requirement or two defensible fixes. Do not ask about something you can
+decide by reading more code.
+
+Report the single most important real issue in this form:
+
+```
+1. <path>:<line>
+2. issue: <one sentence: what is wrong and why it matters>
+3. fix: <one line: the concrete change>
+4. severity: high | medium | low
+```
+
+If there is no real issue, report `NO ISSUE` and one line saying why. A clean review
+is valid. Do not invent an issue. Use high for a wrong result, data loss, security,
+or a hang or crash on a real path. Use medium for an edge-path bug or a correctness
+risk under load or concurrency. Use low for clarity, a latent foot-gun, or a smell
+with no current failure.
+
+The launcher provides the required bridge reply instructions for every member.

.claude/skills/port-to-opencode/SKILL.md

@@ -53,7 +53,7 @@ Map each entry from `.mcp.json`:
   "$schema": "https://opencode.ai/config.json",
   "instructions": ["CLAUDE.md"],
   "mcp": {
-    "bridged":  { "type": "remote", "url": "http://127.0.0.1:8765/mcp", "enabled": true },
+    "fleetd":  { "type": "remote", "url": "http://127.0.0.1:8765/mcp", "enabled": true },
     "context7": { "type": "remote", "url": "https://example.dev/mcp", "enabled": true,
                   "headers": { "Authorization": "Bearer {env:CONTEXT7_TOKEN}" } },
     "gitea":    { "type": "local", "command": ["gitea-mcp", "-t", "stdio"], "enabled": true,

.opencode/agent/architect.md

@@ -0,0 +1,24 @@
+---
+description: Refine work into clear, independent units before implementation.
+mode: primary
+---
+
+<!-- CB-617: The model comes from bridged.yaml because the launch flag overrides model here on both backends. -->
+
+You are an architect in this fleet. You refine work before anyone builds it: scope,
+acceptance criteria, risks, and a unit split. You read the repo and write analysis.
+You never commit production code and never open a pull request.
+
+A design task is worked by two architects. Design alone first, then exchange and
+say plainly where you disagree. Do not concede just to agree.
+
+Do only the assigned scope. Note anything outside that scope in one line and do not
+investigate it further. Use `bridge_ask{question}` only when a decision belongs to
+the lead, such as an unclear requirement or two defensible fixes. Do not ask about
+something you can decide by reading more code.
+
+Report only work you actually did and the real output of checks you ran. Do not
+claim a result from a tool you could not use. The primary's IDE tools are not yours.
+A mounted forge tool may use a blocked credential and fail by design.
+
+The launcher provides the required bridge reply instructions for every member.

.opencode/agent/dev.md

@@ -0,0 +1,29 @@
+---
+description: Implement one assigned unit, test it, and open a pull request.
+mode: primary
+---
+
+<!-- CB-617: The model comes from bridged.yaml because the launch flag overrides model here on both backends. -->
+
+You implement the one unit you were given and nothing else. Work in your assigned
+git worktree and branch. Never check out, rebase onto, or push to `main`. Confirm
+the worktree root and branch before you edit. Use only paths under that root.
+
+Do only the assigned scope. Note anything outside that scope in one line and do not
+investigate it further. Use `bridge_ask{question}` only when a decision belongs to
+the lead, such as an unclear requirement or two defensible fixes. Do not ask about
+something you can decide by reading more code.
+
+Implement the change and run the full required build in your worktree. Read the
+complete output and report its real result. Do not hide failures with a pipe. State
+only checks you actually ran. The primary's IDE tools are not yours. A mounted forge
+tool may use a blocked credential and fail by design.
+
+Stage only files you changed. Never use `git add -A` or `git add .`. Never commit
+`.mcp.json` or `wiki/`. Commit with a clear message, push your branch, and open your
+own pull request against `main`. Never merge.
+
+Your handoff must name the pull request or why it was not created, the branch, the
+files changed, the build result, and any caveat for review.
+
+The launcher provides the required bridge reply instructions for every member.

.opencode/agent/reviewer.md

@@ -0,0 +1,34 @@
+---
+description: Review one assigned scope and report the most important real issue.
+mode: primary
+---
+
+<!-- CB-617: The model comes from bridged.yaml because the launch flag overrides model here on both backends. -->
+
+You review the diff you were given. Report bugs, risks, and missing tests. You do
+not change code.
+
+Read the whole assigned scope before judging it. Review only that scope. If you see
+something outside it, note it in one line and do not investigate it further. Do not
+run the build. The owner makes changes and runs checks.
+
+Use `bridge_ask{question}` only when a decision belongs to the lead, such as an
+unclear requirement or two defensible fixes. Do not ask about something you can
+decide by reading more code.
+
+Report the single most important real issue in this form:
+
+```
+1. <path>:<line>
+2. issue: <one sentence: what is wrong and why it matters>
+3. fix: <one line: the concrete change>
+4. severity: high | medium | low
+```
+
+If there is no real issue, report `NO ISSUE` and one line saying why. A clean review
+is valid. Do not invent an issue. Use high for a wrong result, data loss, security,
+or a hang or crash on a real path. Use medium for an edge-path bug or a correctness
+risk under load or concurrency. Use low for clarity, a latent foot-gun, or a smell
+with no current failure.
+
+The launcher provides the required bridge reply instructions for every member.

CLAUDE.md

@@ -78,7 +78,7 @@ below are the procedure — run them in order, every task, not only the big ones
    what makes them reliable. Where the project ships no such skill, spell the procedure out in the
    brief instead. The brief is self-contained — the worker sees your message and the repo, nothing
    of your context, your plan, or your screen.
-5. **Collect** — `bridge_poll{ticket}` → `bridge_ack{ticket, msgId}`. Answer a worker's `bridge_ask`
+5. **Collect** — `bridge_poll{ticket}` → `bridge_ack{target, msgId}`. Answer a worker's `bridge_ask`
    with `bridge_send{turnId, content}` — **not** `sessionId`. A worker gone quiet is diagnosed with
    `bridge_status`, never by reading its terminal; it also reports an open question and the `turnId`
    that answers it. **A worker's ask waits ~55 seconds, and no nudge makes that longer** — so never
@@ -173,12 +173,14 @@ you.
 |---|---|---|
 | the launcher's reply charter | the one rule that must survive with no repo: *end every turn with `bridge_reply`* | every spawned member, at launch, every peer kind — never a lead |
 | **this section** | protocol + orchestration policy | primary **and** every member that reads the repo — tracked in git, so worktrees inherit it |
+| role agent definition files | role contract and per-job procedure | a member whose launcher binds its role to the matching file in its worktree |
 | role playbook skills | per-job procedure (commit/PR recipe, finding format) | a member told to load one |
 | the bridge's own docs | design detail, flows, error model | on demand |
 
-A rule belongs in **exactly one** layer — the outermost one that must obey it. Peers that don't read
-`CLAUDE.md` (non-Claude adapters) get the charter only, so any rule *they* must obey belongs in the
-charter, not here.
+A rule belongs in **exactly one** layer — the outermost one that must obey it. A member without a
+repo checkout still gets the launcher's reply charter, which is why that one rule stays there.
+Peers that don't read `CLAUDE.md` (non-Claude adapters) get the charter only, so any rule *they*
+must obey belongs in the charter, not here.
 
 ## Project addendum — claude-bridge (not part of the canonical block)
 
@@ -192,8 +194,10 @@ charter, not here.
 - **Never commit** `.mcp.json` (the primary's local copy, flagged `--skip-worktree`) or `wiki/`
   (a submodule with its own remote).
 - **Flows and the error model** — rendezvous, `bridge_ask`, detached delivery, turn-done fallback —
-  are diagrammed in `docs/MCP-Contract.md` §6, kept out of this file because it loads into every
-  session's context.
+  are diagrammed in `docs/MCP-Contract.md` **§6 only**. The rest of that page is a pre-build design
+  doc whose tool names, parameter names and REST paths never caught up with the code, so do not use
+  it as the tool reference (CB-609). Section 6 is kept out of this file because this file loads into
+  every session's context.
 
 ### Redeploying the daemon — the lead may do this (primary only)
 

README.md

@@ -18,8 +18,9 @@ one unified Claude setup and the **sole communication gateway** (REST/SSE stays
 clients; any broker is `bridged`-internal, below the gateway).
 herdr owns the PTYs, multiplexing, persistence, and **agent-status events**; `bridged` owns
 policy (subscription boundary, session lifecycle, status-gated delivery) and the client
-contract. The worker `claude` launches with `ANTHROPIC_BASE_URL=https://ollama.ltms.dev` + a
-bearer token; the primary Opus stays env-clean and calls `bridged`'s MCP tools.
+contract. A Claude member launches with `ANTHROPIC_BASE_URL` pointed at the gateway,
+`https://llm.ltms.dev/anthropic`, plus a bearer token; the lead stays env-clean and calls
+`bridged`'s MCP tools. See the wiki's **[13 User Guide](wiki/13-User-Guide.md)** to run it.
 
 ```mermaid
 flowchart LR
@@ -31,7 +32,7 @@ flowchart LR
     end
     HERDR["herdr<br/>panes · agent-status"]
     W["worker claude pane<br/>ANTHROPIC_BASE_URL set<br/>MCP client"]
-    M["ollama.ltms.dev<br/>(worker model)"]
+    M["llm.ltms.dev<br/>(the one gateway)"]
 
     OPUS -->|"MCP bridge_send (blocks)"| SRV
     W -.->|"MCP bridge_reply"| SRV

bridged/bridged.example.yaml

@@ -35,7 +35,7 @@ bind:
 # a caller whose connection maps to this pane resolves as the primary (no credential needed —
 # the pane mapping is as unforgeable as a worker's), and reply nudges are pushed to it.
 # REQUIRED when the primary runs inside a herdr pane — without it the pane match reads the
-# primary as a worker and refuses spawn/send/stop. Get the id from bridge_whoami; re-pin if
+# primary as a worker and refuses spawn/send/stop. Get the id from fleet_whoami; re-pin if
 # the primary moves panes.
 # primary:
 #   terminal: term_0123456789abcd
@@ -49,12 +49,12 @@ bind:
 #
 #   tab → the ONLY field identity depends on (CB-579); the exact label of the tab hosting the lead.
 #         Label the tab yourself, or let bridged label one it launches — see `fleet.leaders:` below.
-#   kind/model → descriptive; they document what runs in the pane and are echoed by bridge_whoami
+#   kind/model → descriptive; they document what runs in the pane and are echoed by fleet_whoami
 #
 # A lead's tab must already carry its label (or be launched by bridged, which labels it) — there is
 # no terminal id to paste in and nothing to re-pin when the session restarts: the tab survives, so
 # the same label resolves the same lead again on the next scan.
-# `bridge_whoami` reports `{"role":"primary","leader":"<name>"}`; role stays "primary" because a lead
+# `fleet_whoami` reports `{"role":"primary","leader":"<name>"}`; role stays "primary" because a lead
 # IS a primary for authorization, so nothing that keys on the role breaks.
 #
 # KEEP `primary:` when adding leads: it still addresses the CB-307 push loop, which needs a single
@@ -69,7 +69,7 @@ bind:
 # never a capability: what a pane may do is decided by the role the daemon resolves for it.
 
 # CB-551: IDLE-LEAD HEARTBEAT — nudge the single lead back to work when it has been continuously
-# idle (no open bridge_send driving it) past the quiet period. The fleet is one lead + architects +
+# idle (no open fleet_send driving it) past the quiet period. The fleet is one lead + architects +
 # workers, so a lead that stalls is a single point of failure; the ReplyPushLoop only nudges when a
 # reply lands, and this timer catches the gap where nothing lands and the lead just sits idle.
 #
@@ -99,7 +99,7 @@ bind:
 #                      them changes nothing right now, and no minimum is enforced on either, because
 #                      nothing reads them to enforce one. They exist so a later build can start
 #                      honouring them without another config-shape change.
-#   notifications.mode → "webhook" flips what bridge_list REPORTS (healthCoverage: "full" instead
+#   notifications.mode → "webhook" flips what fleet_list REPORTS (healthCoverage: "full" instead
 #                      of "detection-only") — it does NOT make bridged send any webhook call; no
 #                      delivery mechanism is implemented yet. Any other value, or omitting the
 #                      block, reports "detection-only".
@@ -152,7 +152,7 @@ herdrSocket: ~/.config/herdr/herdr.sock
 #   gitHostEnv       → host env var holding the forge host (default GITEA_HOST). Injected as
 #                      GITEA_HOST *only* alongside a resolved gitTokenEnv.
 #   exhaustedPattern → regex matched against a completion-fallback scrape (CB-578 stage A) to
-#                      classify a turn that ended with no bridge_reply as the backend having
+#                      classify a turn that ended with no fleet_reply as the backend having
 #                      refused on a subscription usage limit, rather than a real answer. Opt-in —
 #                      omit and this profile's completion fallback behaves exactly as before.
 #                      Every backend words its refusal differently, so this is config, never a
@@ -200,14 +200,37 @@ profiles:
     # weight: relative selection weight for automatic placement (weighted, round-robin, and
     # fixed's fallback walk). Absent defaults to 1.0. An explicit 0 or negative value means
     # "never auto-select this profile" (CB-554) — it stays reachable via an explicit
-    # `bridge_spawn{profile:"gx10"}`, which bypasses placement entirely; only automatic
+    # `fleet_spawn{profile:"gx10"}`, which bypasses placement entirely; only automatic
     # selection skips it.
     weight: 0.5
     # maxLoad: max live workers on this profile. Omit for unlimited. An explicit 0 (CB-585) caps
     # the profile at zero live members — it is excluded from automatic placement and an explicit
-    # `bridge_spawn{profile:"gx10"}` against it is refused too; a cap holds even when the profile
+    # `fleet_spawn{profile:"gx10"}` against it is refused too; a cap holds even when the profile
     # is named directly. Negative is refused at config load — there is no sane meaning for it.
     maxLoad: 2
+    # subscription: true
+    #   THE KNOB THAT DECIDES WHO PAYS (CB-539). Default false. When true, this profile's members
+    #   run on the OPERATOR'S OWN Claude subscription instead of a metered endpoint — every spawn
+    #   bills your plan and eats your usage limit. Off-subscription is the whole point of this
+    #   daemon, so treat `true` as a deliberate exception, not a convenience.
+    #
+    #   What changes when it is set (ClaudeCodeLauncher):
+    #     - no ANTHROPIC_BASE_URL and no ANTHROPIC_AUTH_TOKEN are injected — the member inherits
+    #       the operator's own Claude Code auth, which is exactly why it bills the plan;
+    #     - SubscriptionGuard never vets it, because there is no baseUrl to vet;
+    #     - no token is required, so `tokenEnv` is irrelevant here.
+    #
+    #   MUTUALLY EXCLUSIVE with `baseUrl` — setting both is refused at config load (CB-542). On the
+    #   subscription path no guard would vet the URL, so allowing both would be a way around the
+    #   guard rather than a configuration.
+    #
+    #   GOTCHA 1 — it is invisible to the startup secret check. `Bridged.reportRequiredSecrets`
+    #   skips subscription profiles on purpose (they need no token), so a boot log that reports
+    #   every secret as fine says nothing about these profiles.
+    #
+    #   GOTCHA 2 — `maxLoad` is the ONLY throttle you have here. There is no metering, no budget
+    #   and no refusal on cost; the cap on live members is the single thing standing between a
+    #   fan-out and your monthly limit. Set it deliberately and keep it small.
     # gitTokenEnv: GITEA_TOKEN        # opt-in: let this profile's workers open their own PR (CB-302)
     # gitHostEnv: GITEA_HOST          # defaults to GITEA_HOST; injected only with gitTokenEnv
     # exhaustedPattern: "usage limit has been reached"   # opt-in: classify a usage-limit refusal (CB-578)
@@ -236,8 +259,8 @@ profiles:
   # The bridge MCP + reply charter mount via a generated OPENCODE_CONFIG file, and the model is a
   # `provider/model` selector. Placement, tabs, cwd, and the readiness gate are shared with Claude.
   #
-  # Dogfood-verified 2026-07-29 against opencode 1.18.5 (spawn → readiness gate → bridge_send →
-  # structured bridge_reply → teardown). The `opencode/*-free` models run on opencode's own gateway
+  # Dogfood-verified 2026-07-29 against opencode 1.18.5 (spawn → readiness gate → fleet_send →
+  # structured fleet_reply → teardown). The `opencode/*-free` models run on opencode's own gateway
   # and need NO credentials — check `opencode models` for the current free list, since the names
   # change. That also makes the worker off-subscription by construction.
   #   opencode-free:
@@ -424,7 +447,7 @@ fleet:
   #                                  # MUST NOT be a member workspace — those are excluded from the
   #                                  # scan, so a lead placed in one is never found again.
   #     cwd: /path/to/repo           # the launched lead's working directory (default: bridged's own)
-  #     kind: claude                 # descriptive; reported by bridge_whoami
+  #     kind: claude                 # descriptive; reported by fleet_whoami
   #   gpt-sol-5.6:
   #     tab: "lead: gpt-sol-5.6"
   #     kind: opencode
@@ -577,7 +600,7 @@ guard:
 #   uri: amqp://guest:guest@127.0.0.1:5672
 #   prefetch: 32
 
-# Active push-to-primary (CB-307 Stage 3). When a worker reply lands with no open bridge_send,
+# Active push-to-primary (CB-307 Stage 3). When a worker reply lands with no open fleet_send,
 # the ReplyPushLoop injects a *drain nudge* (never the payload) into the primary's own herdr
 # pane — status-gated (only when injectable, never mid-turn) and bounded. Ack = drain: the loop
 # stops as soon as the primary's inbox is empty.
@@ -592,7 +615,7 @@ guard:
 #                    classified as a WORKER and refused SPAWN/SEND/STOP. That failure is
 #                    self-locking: the learned terminal is populated by the very orchestration
 #                    calls being refused, so only this pinned value can break the cycle. Read the
-#                    id off bridge_whoami (it reports the current terminal even while
+#                    id off fleet_whoami (it reports the current terminal even while
 #                    misclassified) and re-pin whenever the primary moves panes.
 #   pushReminders  → max nudges before giving up (default 5)
 #   pushBackoffMs  → delay between nudges in ms (default 15000)

bridged/src/main/java/dev/ltms/bridged/member/ClaudeCodeLauncher.java

@@ -9,6 +9,10 @@ import dev.ltms.bridged.peer.Capability;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
+import java.io.IOException;
+import java.io.UncheckedIOException;
+import java.nio.file.Files;
+import java.nio.file.Path;
 import java.util.EnumSet;
 import java.util.List;
 import java.util.Map;
@@ -230,7 +234,7 @@ public final class ClaudeCodeLauncher extends HerdrPeerLauncher {
         // model flag so --model keeps outranking the operator's own argv.
         // mutableArgv: argvWithBridge may hand back the profile's own (immutable) List.of when it
         // has neither MCP nor a charter — session flags must be added into a list we own.
-        List<String> argv = mutableArgv(argvWithBridge(cfg, spec.charter()));
+        List<String> argv = mutableArgv(argvWithBridge(cfg, spec));
         String agentSessionId = applySessionIdentity(argv, spec.sessionName(), spec.resumeSessionId());
         return new Launch(workerEnv, argvWithModel(argv, cfg), agentSessionId);
     }
@@ -265,13 +269,31 @@ public final class ClaudeCodeLauncher extends HerdrPeerLauncher {
     }
 
     /**
-     * The launch argv, plus an inline {@code --mcp-config} when {@code worker.mcpUrl} is set and
-     * {@code --append-system-prompt} when the base composed a charter. Neither touches the profile's
-     * config; both are pure command-line flags. This inline-flag mount is Claude Code specific —
-     * other adapters mount MCP and instructions their own way.
+     * The launch argv, plus an inline {@code --mcp-config} when {@code worker.mcpUrl} is set, the
+     * CB-617 charter flags, and {@code --agent <role>} when the role has an agent-definition file
+     * under the worker's cwd. Neither touches the profile's config; all are pure command-line flags.
+     * This inline-flag mount is Claude Code specific — other adapters mount MCP and instructions
+     * their own way.
+     *
+     * <p>CB-617: the role charter is operator-authored and often multi-line, so it can never be a
+     * single inline argv element — herdr refuses to shell-encode a multi-line argument
+     * ({@code invalid_agent_argument}). It is written to a temp file instead and mounted with
+     * {@code --append-system-prompt-file}, which this host confirms Claude Code accepts for a
+     * multi-line file.
+     *
+     * <p>CB-618: Claude Code refuses to start when BOTH {@code --append-system-prompt} and
+     * {@code --append-system-prompt-file} are on the command line ("Cannot use both ... Please use
+     * only one"), so the two charters can never travel on separate flags. When both are present they
+     * are concatenated into the one file, role charter first and reply charter last — last is where
+     * the reply rule must sit, because it is the rule that must survive. When only the reply charter
+     * is present it keeps its proven inline {@code --append-system-prompt} delivery, which is also
+     * the only form that reaches a member with no repo checkout.
      */
-    private List<String> argvWithBridge(BridgedConfig.Profile cfg, String charter) {
-        if (!cfg.hasMcp() && charter == null) {
+    private List<String> argvWithBridge(BridgedConfig.Profile cfg, LaunchSpec spec) {
+        String roleCharter = nonBlank(spec.roleCharter());
+        String replyCharter = nonBlank(spec.replyCharter());
+        Path agentFile = agentDefinitionFile(spec.cwd(), spec.role(), ".claude", "agents");
+        if (!cfg.hasMcp() && roleCharter == null && replyCharter == null && agentFile == null) {
             return cfg.argv();
         }
         List<String> argv = mutableArgv(cfg.argv());
@@ -281,13 +303,45 @@ public final class ClaudeCodeLauncher extends HerdrPeerLauncher {
             argv.add("--mcp-config");
             argv.add(mcpJson);
         }
-        if (charter != null) {
+        if (roleCharter != null) {
+            String combined = replyCharter == null ? roleCharter : roleCharter + "\n\n" + replyCharter;
+            argv.add("--append-system-prompt-file");
+            argv.add(writeCharterFile(combined).toString());
+        } else if (replyCharter != null) {
             argv.add("--append-system-prompt");
-            argv.add(charter);
+            argv.add(replyCharter);
+        }
+        if (agentFile != null) {
+            argv.add("--agent");
+            argv.add(spec.role().wireName());
         }
         return argv;
     }
 
+    /** {@code s}, or {@code null} when {@code s} is null/blank — the charter-presence test used above. */
+    private static String nonBlank(String s) {
+        return (s == null || s.isBlank()) ? null : s;
+    }
+
+    /**
+     * Write the role charter to a fresh temp file so it can be mounted with
+     * {@code --append-system-prompt-file} instead of riding inline in argv (CB-617). Best-effort
+     * cleaned via {@code deleteOnExit} — the same disposable-worker-config cleanup
+     * {@link OpenCodeLauncher#writeConfig} already uses for its charter file, since the process that
+     * reads this file (the spawned peer) outlives this JVM call and there is no spawn-scoped teardown
+     * hook to delete it synchronously.
+     */
+    private static Path writeCharterFile(String charterText) {
+        try {
+            Path file = Files.createTempFile("bridged-role-charter-", ".md");
+            Files.writeString(file, charterText);
+            file.toFile().deleteOnExit();
+            return file;
+        } catch (IOException e) {
+            throw new UncheckedIOException("cannot write role charter temp file", e);
+        }
+    }
+
     /**
      * Pin the model on the command line as well as in {@code ANTHROPIC_MODEL} (CB-533).
      *

bridged/src/main/java/dev/ltms/bridged/member/HerdrPeerLauncher.java

@@ -17,6 +17,8 @@ import dev.ltms.bridged.peer.SpawnRequest;
 import org.slf4j.Logger;
 import org.slf4j.LoggerFactory;
 
+import java.nio.file.Files;
+import java.nio.file.Path;
 import java.security.SecureRandom;
 import java.util.ArrayList;
 import java.util.Collection;
@@ -279,8 +281,35 @@ public abstract class HerdrPeerLauncher implements PeerLauncher {
         }
     }
 
-    /** All per-spawn values adapters may need, including the base-composed effective charter. */
-    protected record LaunchSpec(String sessionName, String resumeSessionId, MemberRole role, String charter) {
+    /**
+     * All per-spawn values adapters may need.
+     *
+     * <p>{@code charter} is the base-composed effective charter (role charter, then the reply
+     * charter, joined by a blank line) — kept for an adapter that mounts both as one blob (opencode
+     * writes it to a single file) and as the exact input {@link CharterReceipt#compose} fingerprints.
+     * {@code roleCharter} and {@code replyCharter} are the same text split back into its two parts
+     * (CB-617), for an adapter that must deliver them differently: the role charter is
+     * operator-authored and often multi-line, so it cannot travel as an inline argv element (herdr
+     * refuses to shell-encode a multi-line argument); the reply charter is always one line and is
+     * demonstrated to encode, so it may still go inline. {@code cwd} is the spawn's resolved working
+     * directory (CB-112), needed to look up a role's agent-definition file before the peer starts.
+     */
+    protected record LaunchSpec(String sessionName, String resumeSessionId, MemberRole role, String charter,
+                                String roleCharter, String replyCharter, String cwd) {
+    }
+
+    /**
+     * The role's agent-definition file under {@code <cwd>/<dir1>/<dir2>/<role>.md}, or {@code null}
+     * when absent or inapplicable (no role, no cwd, or the file does not exist) — CB-617. A member
+     * whose role has no such file must still spawn, so this is a lookup, never a requirement: the
+     * caller passes {@code --agent <role>} only when the return value is non-null.
+     */
+    protected static Path agentDefinitionFile(String cwd, MemberRole role, String dir1, String dir2) {
+        if (role == null || cwd == null || cwd.isBlank()) {
+            return null;
+        }
+        Path candidate = Path.of(cwd, dir1, dir2, role.wireName() + ".md");
+        return Files.isRegularFile(candidate) ? candidate : null;
     }
 
     // --- profile surface -----------------------------------------------------------------------
@@ -387,9 +416,10 @@ public abstract class HerdrPeerLauncher implements PeerLauncher {
         // start has no bridge_spawn result and no roster row, so the failure log below is the only
         // surface the byte count can appear on. The charter text itself is never logged.
         CharterReceipt receipt = CharterReceipt.compose(role, cfg.profile(), roleCharter, charter);
+        String cwd = resolveCwd(requestedCwd, cfg, callerCwd);
         try {
-            Launch launch = buildLaunch(cfg, new LaunchSpec(sessionName, resumeSessionId, role, charter));
-            String cwd = resolveCwd(requestedCwd, cfg, callerCwd);
+            Launch launch = buildLaunch(cfg, new LaunchSpec(sessionName, resumeSessionId, role, charter,
+                    roleCharter, replyCharter, cwd));
             Agent agent = cfg.tabPlacement()
                     ? spawnInTab(cfg, launch.env(), launch.argv(), cwd, role, liveFleet)
                     : spawnAsPane(cfg, launch.env(), launch.argv(), cwd, charter);

bridged/src/main/java/dev/ltms/bridged/member/OpenCodeLauncher.java

@@ -208,8 +208,24 @@ public final class OpenCodeLauncher extends HerdrPeerLauncher {
             workerEnv.put("OPENCODE_CONFIG", writeConfig(cfg, spec.charter()).toString());
         }
         applyGitToken(workerEnv, cfg);
-        return new Launch(workerEnv,
-                argvWithResume(argvWithModel(argvWithAuto(cfg), cfg), spec.resumeSessionId()));
+        List<String> argv = argvWithResume(argvWithModel(argvWithAuto(cfg), cfg), spec.resumeSessionId());
+        return new Launch(workerEnv, argvWithAgent(argv, spec));
+    }
+
+    /**
+     * The launch argv plus, when the role has an agent-definition file under the worker's cwd,
+     * opencode's {@code --agent <role>} flag (CB-617). A role with no such file gets nothing added —
+     * the member must still spawn.
+     */
+    private List<String> argvWithAgent(List<String> argv, LaunchSpec spec) {
+        Path agentFile = agentDefinitionFile(spec.cwd(), spec.role(), ".opencode", "agent");
+        if (agentFile == null) {
+            return argv;
+        }
+        List<String> withAgent = mutableArgv(argv);
+        withAgent.add("--agent");
+        withAgent.add(spec.role().wireName());
+        return withAgent;
     }
 
     /**

bridged/src/test/java/dev/ltms/bridged/member/ClaudeCodeLauncherTest.java

@@ -15,8 +15,11 @@ import dev.ltms.bridged.peer.PeerHandle;
 import dev.ltms.bridged.peer.PeerUnreachableException;
 import dev.ltms.bridged.peer.SpawnRequest;
 import org.junit.jupiter.api.Test;
+import org.junit.jupiter.api.io.TempDir;
 import org.slf4j.LoggerFactory;
 
+import java.nio.file.Files;
+import java.nio.file.Path;
 import java.util.List;
 import java.util.Map;
 import java.util.Set;
@@ -106,10 +109,24 @@ class ClaudeCodeLauncherTest {
                 "the operator's own args are preserved, in order, ahead of the model flag");
     }
 
+    // --- CB-617/CB-618: both charters travel in ONE file; the reply charter alone stays inline ---
+
+    /**
+     * CB-617: herdr refuses to shell-encode a multi-line inline argv argument
+     * ({@code invalid_agent_argument}) — the exact failure this reproduced on profile {@code opus}.
+     * The role charter is operator-authored and often multi-line, so it must never appear as an argv
+     * element.
+     *
+     * <p>CB-618: and Claude Code itself refuses to start when both {@code --append-system-prompt} and
+     * {@code --append-system-prompt-file} are given ("Cannot use both ... Please use only one"), so
+     * the reply charter cannot ride inline alongside a role charter either. Both go in the one file,
+     * reply charter last. This drives the real launcher entry point ({@code spawn}), the same path a
+     * live spawn takes — not the argv builder in isolation.
+     */
     @Test
-    void appendsTheBaseComposedRoleAndReplyCharter() {
+    void bothChartersTravelInOneFileAndNeverOnBothFlags() {
         FakeHerdr herdr = new FakeHerdr();
-        String roleCharter = "You review changes.";
+        String roleCharter = "You review changes.\nLine two.\nLine three.";
         BridgedConfig.Profile cfg = new BridgedConfig.Profile(
                 "sonnet", "http://gx00.gw:8000", null, null, "BRIDGED_WORKER_TOKEN",
                 List.of("claude"), "tab", "bridged-workers", "w #{n}",
@@ -121,11 +138,20 @@ class ClaudeCodeLauncherTest {
         svc.spawn(new SpawnRequest("sonnet", null, null, null, null, MemberRole.REVIEWER));
 
         List<String> args = spawnedArgs(herdr);
-        int flag = args.indexOf("--append-system-prompt");
-        assertEquals(1, args.stream().filter("--append-system-prompt"::equals).count(),
-                "the composed charter is passed once");
-        assertEquals(roleCharter + "\n\n" + HerdrPeerLauncher.REPLY_CHARTER, args.get(flag + 1),
-                "the role charter comes first and the reply rule comes last");
+        assertTrue(args.stream().noneMatch(a -> a.contains("\n")),
+                "no argv element may be multi-line — herdr cannot shell-encode one: " + args);
+
+        int fileFlag = args.indexOf("--append-system-prompt-file");
+        assertTrue(fileFlag >= 0, "the role charter is mounted via --append-system-prompt-file: " + args);
+        assertFalse(args.contains("--append-system-prompt"),
+                "CB-618: Claude Code refuses to start with both flags — the reply charter must not "
+                        + "ride inline beside a role charter: " + args);
+        assertDoesNotThrow(() -> {
+            String written = Files.readString(Path.of(args.get(fileFlag + 1)));
+            assertTrue(written.startsWith(roleCharter), "the file opens with the role charter: " + written);
+            assertTrue(written.endsWith(HerdrPeerLauncher.REPLY_CHARTER),
+                    "the reply charter is last — it is the rule that must survive: " + written);
+        }, "the --append-system-prompt-file path must be a readable file");
     }
 
     @Test
@@ -135,11 +161,13 @@ class ClaudeCodeLauncherTest {
 
         svc.spawn(new SpawnRequest("sonnet", null, null, null, null, MemberRole.DEV));
 
-        assertFalse(spawnedArgs(herdr).contains("--append-system-prompt"));
+        List<String> args = spawnedArgs(herdr);
+        assertFalse(args.contains("--append-system-prompt"));
+        assertFalse(args.contains("--append-system-prompt-file"));
     }
 
     @Test
-    void profileWithoutMcpStillGetsItsRoleCharter() {
+    void profileWithoutMcpStillGetsItsRoleCharterAsAFileNotInline() {
         FakeHerdr herdr = new FakeHerdr();
         String roleCharter = "You design changes.";
         ClaudeCodeLauncher svc = labelService(herdr, () -> fleet(Map.of("architect", roleCharter), null));
@@ -147,12 +175,52 @@ class ClaudeCodeLauncherTest {
         svc.spawn(new SpawnRequest("sonnet", null, null, null, null, MemberRole.ARCHITECT));
 
         List<String> args = spawnedArgs(herdr);
-        int flag = args.indexOf("--append-system-prompt");
+        int flag = args.indexOf("--append-system-prompt-file");
         assertTrue(flag >= 0, "a role charter does not need an MCP mount");
-        assertEquals(roleCharter, args.get(flag + 1));
+        assertDoesNotThrow(() -> assertEquals(roleCharter,
+                        Files.readString(Path.of(args.get(flag + 1)))),
+                "the file holds the role charter");
+        assertFalse(args.contains("--append-system-prompt"), "no reply charter without an MCP mount");
         assertFalse(args.contains("--mcp-config"));
     }
 
+    // --- CB-617: --agent <role> when the role has an agent-definition file --------------------
+
+    @Test
+    void agentFlagIsPassedWhenTheRoleAgentDefinitionFileExists(@TempDir Path cwd) throws Exception {
+        FakeHerdr herdr = new FakeHerdr();
+        Path agentsDir = Files.createDirectories(cwd.resolve(".claude/agents"));
+        Files.writeString(agentsDir.resolve("reviewer.md"), "---\nname: reviewer\n---\nBe a reviewer.");
+        BridgedConfig.Profile cfg = new BridgedConfig.Profile(
+                "sonnet", "http://gx00.gw:8000", null, null, "BRIDGED_WORKER_TOKEN",
+                List.of("claude"), "tab", "bridged-workers", "w #{n}", null, null, null);
+        ClaudeCodeLauncher svc = new ClaudeCodeLauncher(new AgentControl(herdr), new WorkspaceControl(herdr),
+                new SubscriptionGuard(Set.of("gx00.gw")), Map.of(cfg.profile(), cfg), cfg.profile(), _ -> null);
+
+        svc.spawn(new SpawnRequest("sonnet", cwd.toString(), null, null, null, MemberRole.REVIEWER));
+
+        List<String> args = spawnedArgs(herdr);
+        int flag = args.indexOf("--agent");
+        assertTrue(flag >= 0, "--agent is passed when the role's agent file exists: " + args);
+        assertEquals("reviewer", args.get(flag + 1));
+    }
+
+    @Test
+    void noAgentFlagWhenTheRoleAgentDefinitionFileIsAbsent(@TempDir Path cwd) {
+        FakeHerdr herdr = new FakeHerdr();
+        BridgedConfig.Profile cfg = new BridgedConfig.Profile(
+                "sonnet", "http://gx00.gw:8000", null, null, "BRIDGED_WORKER_TOKEN",
+                List.of("claude"), "tab", "bridged-workers", "w #{n}", null, null, null);
+        ClaudeCodeLauncher svc = new ClaudeCodeLauncher(new AgentControl(herdr), new WorkspaceControl(herdr),
+                new SubscriptionGuard(Set.of("gx00.gw")), Map.of(cfg.profile(), cfg), cfg.profile(), _ -> null);
+
+        PeerHandle handle = svc.spawn(new SpawnRequest("sonnet", cwd.toString(), null, null, null, MemberRole.REVIEWER));
+
+        assertNotNull(handle, "the member still spawns with no agent-definition file");
+        assertFalse(spawnedArgs(herdr).contains("--agent"),
+                "no --agent flag when the role has no agent-definition file");
+    }
+
     private ClaudeCodeLauncher multiProfile(FakeHerdr herdr) {
         BridgedConfig.Profile gx10 = new BridgedConfig.Profile("gx10", "http://gx10.gw:8000", "coder",
                 null, "BRIDGED_WORKER_TOKEN", List.of("claude"), "tab", "bridged-workers", "w #{n}", null, null, null);

bridged/src/test/java/dev/ltms/bridged/member/OpenCodeLauncherTest.java

@@ -8,6 +8,7 @@ import dev.ltms.bridged.herdr.FakeHerdr;
 import dev.ltms.bridged.herdr.WorkspaceControl;
 import dev.ltms.bridged.peer.Capability;
 import dev.ltms.bridged.peer.CharterReceipt;
+import dev.ltms.bridged.peer.MemberRole;
 import dev.ltms.bridged.peer.PeerHandle;
 import dev.ltms.bridged.peer.PeerUnreachableException;
 import dev.ltms.bridged.peer.SpawnRequest;
@@ -208,6 +209,35 @@ class OpenCodeLauncherTest {
                 "--auto is unconditional: a model-less worker still must never block on approval");
     }
 
+    // --- CB-617: --agent <role> when the role has an agent-definition file --------------------
+
+    @Test
+    void agentFlagIsPassedWhenTheRoleAgentDefinitionFileExists(@TempDir Path root) throws Exception {
+        FakeHerdr herdr = new FakeHerdr();
+        Path agentsDir = Files.createDirectories(root.resolve(".opencode/agent"));
+        Files.writeString(agentsDir.resolve("dev.md"), "You are a dev.");
+        OpenCodeLauncher svc = service(herdr, root, opencodeCfg(null, null, null));
+
+        svc.spawn(new SpawnRequest(null, root.toString(), null, null, null, MemberRole.DEV));
+
+        List<String> args = startArgs(herdr);
+        int flag = args.indexOf("--agent");
+        assertTrue(flag >= 0, "--agent is passed when the role's agent file exists: " + args);
+        assertEquals("dev", args.get(flag + 1));
+    }
+
+    @Test
+    void noAgentFlagWhenTheRoleAgentDefinitionFileIsAbsent(@TempDir Path root) {
+        FakeHerdr herdr = new FakeHerdr();
+        OpenCodeLauncher svc = service(herdr, root, opencodeCfg(null, null, null));
+
+        PeerHandle handle = svc.spawn(new SpawnRequest(null, root.toString(), null, null, null, MemberRole.DEV));
+
+        assertNotNull(handle, "the member still spawns with no agent-definition file");
+        assertFalse(startArgs(herdr).contains("--agent"),
+                "no --agent flag when the role has no agent-definition file");
+    }
+
     @Test
     void injectsForgeTokenWhenProfileGrantsIt(@TempDir Path root) {
         FakeHerdr herdr = new FakeHerdr();

docs/MCP-Contract.md

@@ -1,9 +1,24 @@
 # MCP Contract — `bridged`'s unified gateway
 
-> **Status:** 🟡 Design (2026-07-14). Greenfield — no MCP code exists yet; the pom carries
-> only Javalin/Jackson. This page defines the tool surface that CB-104 and its followers
-> implement. It supersedes nothing; it fills the "MCP server face" left open by the
-> [Architecture](1-Architecture) page.
+> **Status: 🔴 HISTORICAL DESIGN — do NOT use as the tool reference.** Written 2026-07-14, before
+> any MCP code existed. The system shipped and this page never caught up, so **its tool names,
+> parameter names and REST paths are wrong today**. Audited 2026-08-17; the specific drift:
+>
+> - **Tools it names that do not exist:** `bridge_read`, `bridge_cancel`.
+> - **Shipped tools it omits:** `bridge_poll`, `bridge_ack`, `bridge_profiles`, `bridge_whoami`.
+> - **Parameter names are wrong nearly everywhere** — it says `message`/`target`/`timeout_seconds`/
+>   `block` where the code takes `content`/`sessionId`/`timeoutMs`/`wait`; `text` where
+>   `bridge_reply` takes `content`; `target` where `bridge_stop` takes `paneId`.
+> - **REST paths are wrong:** it says `POST /workers` and `DELETE /workers/{paneId}`; the daemon
+>   serves `POST /members` and `DELETE /members/{paneId}`.
+>
+> **The authoritative tool surface is the live MCP schema** (each tool's own description and
+> parameters, as mounted), with the intent→tool table in `CLAUDE.md` as the short form. Both were
+> checked against `mcp/BridgeMcp.java` on 2026-08-17 and are accurate.
+>
+> What is still worth reading here is **§6 — the flows and the error model** (rendezvous,
+> `bridge_ask`, detached delivery, the turn-done fallback). The shapes it describes are the ones
+> that shipped; only the names around them drifted. Rewriting this page is tracked as **CB-609**.
 
 `bridged` is the **sole communication gateway** for every Claude session in the bridge. Both
 the **primary** (Opus, on subscription) and every **worker** (off-subscription Claude Code)

docs/Team.md

@@ -17,7 +17,7 @@ who the workers are, how the lead picks one, and how it runs many at once.
 - **Workers** — a herd of `claude` panes in herdr, each an addressable `bridged` session
   with its **own model/env**:
   - **Claude workers** (clean env, e.g. Sonnet) — reasoning-heavy or high-accuracy subtasks.
-  - **Local workers** (`ANTHROPIC_BASE_URL=https://ollama.ltms.dev`) — bulk, cheap, or
+  - **Local workers** (`ANTHROPIC_BASE_URL=https://llm.ltms.dev/anthropic`) — bulk, cheap, or
     embarrassingly parallel subtasks.
 
 Every worker is still a *real Claude Code process* (inherits `CLAUDE.md`, hooks, skills,
@@ -35,7 +35,7 @@ flowchart TB
     WL1["w-local-1<br/>ANTHROPIC_BASE_URL set"]
     WL2["w-local-2<br/>ANTHROPIC_BASE_URL set"]
     ANT["api.anthropic.com<br/>(Pro/Max)"]
-    OLL["ollama.ltms.dev<br/>(local model)"]
+    OLL["llm.ltms.dev<br/>(gateway to the local model)"]
 
     LEAD -->|"blocking POST /message (target role)"| BD
     BD -->|"Unix socket · send_text · events.subscribe"| HERDR

e2e/bridge_ask_test.py

@@ -1,16 +1,16 @@
 #!/usr/bin/env python3
-"""Live bridge_ask test — the REVERSE rendezvous (CB-205), watched end to end.
+"""Live fleet_ask test — the REVERSE rendezvous (CB-205), watched end to end.
 
-Every other harness drives the forward path: primary `bridge_send` → worker `bridge_reply`.
+Every other harness drives the forward path: primary `fleet_send` → worker `fleet_reply`.
 This drives the one that runs the other way. A worker is told to pause its delegated turn,
-ask the primary a question via `bridge_ask`, and only finish once it has the answer — so the
+ask the primary a question via `fleet_ask`, and only finish once it has the answer — so the
 turn round-trips primary→worker→primary→worker inside a SINGLE delegation.
 
 The mechanics that only this path exercises:
 
   • a worker's mid-turn question surfacing on the primary's *own* blocked send (Outcome.QUESTION),
   • the `turnId` correlation that lets the primary answer the exact paused turn,
-  • the answer resuming that same turn and the worker's final `bridge_reply` landing on the
+  • the answer resuming that same turn and the worker's final `fleet_reply` landing on the
     re-opened forward waiter (never a stale or cross-wired one).
 
 It is two blocking REST calls, no polling:
@@ -57,14 +57,14 @@ ANSWER_COLOR = "blue"
 
 # A task that CANNOT be completed without asking: the worker is not told which color to choose,
 # only that the primary will name one when asked. So a correct final reply is only reachable by
-# actually calling bridge_ask and using the answer.
+# actually calling fleet_ask and using the answer.
 TASK_PROMPT = (
     "You are a bridge worker in a quick coordination game. You do NOT know which color to pick — "
     "only the primary does. Do exactly this, in order:\n"
-    "1. Call the `bridge_ask` tool with EXACTLY this question: \"PICK A COLOR: red or blue?\"\n"
+    "1. Call the `fleet_ask` tool with EXACTLY this question: \"PICK A COLOR: red or blue?\"\n"
     "2. The primary will answer with one color word. Take that color and uppercase it.\n"
-    "3. Call `bridge_reply` with EXACTLY one line: CHOSEN=<COLOR>  (e.g. CHOSEN=GREEN if told green).\n"
-    "Do not guess a color. Do not call bridge_reply before bridge_ask has returned an answer. "
+    "3. Call `fleet_reply` with EXACTLY one line: CHOSEN=<COLOR>  (e.g. CHOSEN=GREEN if told green).\n"
+    "Do not guess a color. Do not call fleet_reply before fleet_ask has returned an answer. "
     "Do nothing else — no file reads, no other tools."
 )
 
@@ -134,7 +134,7 @@ def run(base, profile, repo, send_timeout, answer_timeout):
     rec.update(tid=tid, pane=pane, spawned=True)
     await_ready(base, tid)
 
-    # 1) Delegate the ask-forcing task. This blocks until the worker calls bridge_ask, at which
+    # 1) Delegate the ask-forcing task. This blocks until the worker calls fleet_ask, at which
     #    point our own send unblocks carrying the question and the turnId to answer on.
     print(f"[{now()}] delegating task (blocks until the worker asks; up to {send_timeout}s)…")
     t0 = time.time()
@@ -159,7 +159,7 @@ def run(base, profile, repo, send_timeout, answer_timeout):
         rec.update(phase="no_turnid", detail="question surfaced without a turnId to answer on")
         return rec
 
-    # 2) Answer on that exact turn. This blocks again until the resumed worker calls bridge_reply.
+    # 2) Answer on that exact turn. This blocks again until the resumed worker calls fleet_reply.
     print(f"[{now()}] answering '{ANSWER_COLOR}' on turn {rec['turnId']} (blocks until reply; up to {answer_timeout}s)…")
     t1 = time.time()
     rec["phase"] = "awaiting_reply"
@@ -188,7 +188,7 @@ def run(base, profile, repo, send_timeout, answer_timeout):
 def grade(rec):
     """PASS only if the worker asked, the turn resumed, and the reply reflects the answer."""
     if rec["phase"] == "no_question":
-        return "NO_ASK", "the worker finished/stalled without ever calling bridge_ask"
+        return "NO_ASK", "the worker finished/stalled without ever calling fleet_ask"
     if rec["phase"] in ("send_error", "answer_error", "spawn"):
         return "ERROR", rec.get("detail") or "transport error before the round-trip completed"
     if rec["phase"] == "no_turnid":
@@ -203,7 +203,7 @@ def grade(rec):
             return "OK", "asked, resumed the same turn, and the reply reflected the primary's answer"
         if reflected:
             return "DEGRADED", f"reply reflected the answer but resolved via {rec['replySource']} " \
-                               "(worker did not call bridge_reply cleanly)"
+                               "(worker did not call fleet_reply cleanly)"
         return "WRONG_ANSWER", f"the worker replied but did not reflect '{ANSWER_COLOR}' — " \
                                f"the answer may not have reached the resumed turn: {rec['reply']!r}"
     return "WEDGE", f"unexpected terminal phase {rec['phase']}: {rec.get('detail')}"
@@ -213,16 +213,16 @@ def write_transcript(out_dir, rec, meta):
     path = out_dir / "bridge_ask_transcript.md"
     g, note = grade(rec)
     with path.open("w") as f:
-        f.write(f"# Live bridge_ask — reverse rendezvous — {datetime.now():%Y-%m-%d %H:%M}\n\n")
+        f.write(f"# Live fleet_ask — reverse rendezvous — {datetime.now():%Y-%m-%d %H:%M}\n\n")
         f.write(f"One worker paused its delegated turn to ask the primary, then resumed with the "
                 f"answer (profile `{meta['profile']}`). Result: **`{g}`**.\n\n")
         f.write("## Round-trip\n\n")
         f.write(f"1. **primary → worker** (delegation): the ask-forcing task.\n")
-        f.write(f"2. **worker → primary** (`bridge_ask`, {rec.get('ask_latency')}s): "
+        f.write(f"2. **worker → primary** (`fleet_ask`, {rec.get('ask_latency')}s): "
                 f"{rec.get('question')!r} — surfaced on the primary's blocked send as a "
                 f"`question` with `turnId={rec.get('turnId')}`.\n")
         f.write(f"3. **primary → worker** (answer on that turn): `{ANSWER_COLOR}`.\n")
-        f.write(f"4. **worker → primary** (`bridge_reply`, {rec.get('answer_latency')}s, "
+        f.write(f"4. **worker → primary** (`fleet_reply`, {rec.get('answer_latency')}s, "
                 f"source={rec.get('replySource')}): {rec.get('reply')!r}\n\n")
         f.write(f"> **{g}:** {note}\n")
         if rec.get("detail"):
@@ -231,7 +231,7 @@ def write_transcript(out_dir, rec, meta):
 
 
 def main():
-    ap = argparse.ArgumentParser(description="Live bridge_ask reverse-rendezvous test (CB-205)")
+    ap = argparse.ArgumentParser(description="Live fleet_ask reverse-rendezvous test (CB-205)")
     ap.add_argument("--base", default="http://127.0.0.1:8765")
     ap.add_argument("--profile", default=None)
     ap.add_argument("--repo", default=str(REPO_ROOT))
@@ -241,7 +241,7 @@ def main():
     ap.add_argument("--keep-worker", action="store_true")
     args = ap.parse_args()
 
-    print(f"[{now()}] live bridge_ask: 1 primary, 1 worker "
+    print(f"[{now()}] live fleet_ask: 1 primary, 1 worker "
           f"(profile={args.profile or 'default'}, repo={args.repo})\n")
 
     rec = {"spawned": False, "pane": None}
@@ -262,7 +262,7 @@ def main():
 
     print()
     print("=" * 72)
-    print("LIVE bridge_ask SUMMARY — reverse rendezvous (CB-205)")
+    print("LIVE fleet_ask SUMMARY — reverse rendezvous (CB-205)")
     print(f"  asked:    {rec.get('question')!r}  (turnId={rec.get('turnId')}, {rec.get('ask_latency')}s)")
     print(f"  answered: {ANSWER_COLOR!r}")
     print(f"  replied:  {rec.get('reply')!r}  (source={rec.get('replySource')}, {rec.get('answer_latency')}s)")

e2e/conversation_sustained_test.py

@@ -24,7 +24,7 @@ Usage:
     --turn-timeout  per-turn max wait, seconds    (default 150)
     --keep-worker   do not stop the worker at the end
 
-Exit code: 0 if every turn in the window resolved via a clean bridge_reply with no channel
+Exit code: 0 if every turn in the window resolved via a clean fleet_reply with no channel
 break; 1 otherwise. A live per-turn log streams to stdout so the run can be watched.
 """
 import argparse
@@ -47,12 +47,12 @@ STEPS = [7, 3, 11, 5, 9, 4, 13, 6, 8, 2]
 RULES = (
     "Let's play a running-total game across several messages. The total starts at 0. "
     "In each message I'll tell you to add a number; keep the running total yourself and "
-    "reply via bridge_reply with ONLY the current total as a plain integer — no words, no "
+    "reply via fleet_reply with ONLY the current total as a plain integer — no words, no "
     "punctuation, just the number. Do not restate the arithmetic. First move: add {step}."
 )
-NEXT = ("Add {step}. Reply via bridge_reply with only the new running total.")
+NEXT = ("Add {step}. Reply via fleet_reply with only the new running total.")
 REANCHOR = ("Let's re-sync — the running total is {total}. Now add {step}. Reply via "
-            "bridge_reply with only the new running total.")
+            "fleet_reply with only the new running total.")
 
 
 def parse_int(reply):
@@ -142,15 +142,15 @@ def main():
     print("=" * 72)
     print(f"SUSTAINED CONVERSATION SUMMARY — 1 primary <-> 1 worker over {dur}s (~{dur/60:.1f} min)")
     print(f"  turns:            {turns}")
-    print(f"  clean bridge_reply exchanges: {oks + drifts}/{turns}   (channel breaks: {breaks})")
+    print(f"  clean fleet_reply exchanges: {oks + drifts}/{turns}   (channel breaks: {breaks})")
     print(f"  arithmetic correct (continuity held): {oks}/{turns}   (drifts: {drifts})")
     print(f"  latency:          avg {avg}s over {turns} turns")
     ok = breaks == 0 and turns >= 2
     if ok and drifts == 0:
-        print("  RESULT: PASS — every turn resolved via bridge_reply and the worker held the "
+        print("  RESULT: PASS — every turn resolved via fleet_reply and the worker held the "
               "running total across the whole window.")
     elif ok:
-        print(f"  RESULT: PASS (channel) — every turn resolved via bridge_reply for the full "
+        print(f"  RESULT: PASS (channel) — every turn resolved via fleet_reply for the full "
               f"window; {drifts} arithmetic drift(s) (worker recovered after re-anchor).")
     else:
         print("  RESULT: FAIL — the channel broke on at least one turn (see CHANNEL BREAK above).")

e2e/conversation_test.py

@@ -4,7 +4,7 @@ the running `bridged` daemon, fully captured, with automatic gap analysis.
 
 This is the repeatable form of the ad-hoc channel test that surfaced the CB-115 gaps
 (herdr `unknown` misclassification, dirty completion scrape, workers not calling
-bridge_reply). It drives a real off-subscription worker over the live gateway exactly
+fleet_reply). It drives a real off-subscription worker over the live gateway exactly
 as a primary Opus session would (async fire-and-poll), records every turn, and grades
 the channel.
 
@@ -130,9 +130,9 @@ def grade(rec):
     phase, source, reply = rec["phase"], rec["source"], rec["reply"]
     has_reply = bool(reply and reply.strip())
     if phase == "done" and source == "reply" and has_reply:
-        return "OK", "clean explicit bridge_reply"
+        return "OK", "clean explicit fleet_reply"
     if phase == "done" and has_reply:
-        return "DEGRADED", f"resolved via {source} (worker did not call bridge_reply)"
+        return "DEGRADED", f"resolved via {source} (worker did not call fleet_reply)"
     if phase == "done" and not has_reply:
         return "EMPTY", "turn completed but reply was empty"
     if phase == "failed":
@@ -206,7 +206,7 @@ def main():
     ok = all(g in ("OK", "DEGRADED") for g in grades)
     reply_clean = all(g == "OK" for g in grades)
     if reply_clean:
-        print("  RESULT: PASS — every turn delivered and got a clean bridge_reply.")
+        print("  RESULT: PASS — every turn delivered and got a clean fleet_reply.")
     elif ok:
         print("  RESULT: PASS (with notes) — every turn delivered & replied, but some via fallback.")
     else:

e2e/issue_hunt_test.py

@@ -12,7 +12,7 @@ and collects every reply concurrently. That stresses what a single worker never
   • reply routing under concurrency (worker A's answer must never resolve worker B's send),
 
 and, as the payload, whether a fleet of off-subscription workers can actually surface
-real issues in the repo and report them back structurally via bridge_reply.
+real issues in the repo and report them back structurally via fleet_reply.
 
 It talks ONLY to the bridge's REST face on loopback — it never sets ANTHROPIC_BASE_URL
 and never touches herdr directly, so it is subscription-safe by construction.
@@ -64,7 +64,7 @@ PROMPT_TMPL = (
     "You are one of several issue-hunting workers in the claude-bridge repo (it is your "
     "current working directory). Your assignment: inspect the file `{target}` and find the "
     "SINGLE most important real bug, correctness gap, or risk in it. Read the file before "
-    "answering. Reply via bridge_reply with EXACTLY these four lines:\n"
+    "answering. Reply via fleet_reply with EXACTLY these four lines:\n"
     "1. {target}:<line>\n"
     "2. issue: <one sentence>\n"
     "3. fix: <one line>\n"
@@ -147,9 +147,9 @@ def grade(rec):
     phase, source, reply = rec["phase"], rec["source"], rec["reply"]
     has_reply = bool(reply and reply.strip())
     if phase == "done" and source == "reply" and has_reply:
-        return "OK", "clean explicit bridge_reply"
+        return "OK", "clean explicit fleet_reply"
     if phase == "done" and has_reply:
-        return "DEGRADED", f"resolved via {source} (worker did not call bridge_reply)"
+        return "DEGRADED", f"resolved via {source} (worker did not call fleet_reply)"
     if phase == "done" and not has_reply:
         return "EMPTY", "turn completed but reply was empty"
     if phase == "failed":

plugin/.mcp.json

@@ -1,6 +1,6 @@
 {
   "mcpServers": {
-    "bridged": {
+    "fleetd": {
       "type": "http",
       "url": "http://127.0.0.1:8765/mcp"
     }

plugin/skills/setup/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: setup
-description: Make the current project bridge-ready — check the prerequisites, mount the bridged MCP gateway into the project's .mcp.json, apply standard Claude Code settings, and verify this session resolves as the primary. Writes no credentials. Load this when asked to set up, install, configure, or onboard a project onto claude-bridge, or when bridge_* tools are expected but absent.
+description: Make the current project bridge-ready — check the prerequisites, mount the bridged MCP gateway into the project's .mcp.json, apply standard Claude Code settings, and verify this session resolves as the primary. Writes no credentials. Load this when asked to set up, install, configure, or onboard a project onto claude-bridge, or when fleet_* tools are expected but absent.
 ---
 
 # Bridge setup — make this project bridge-ready
@@ -70,7 +70,7 @@ The entry to add, exactly:
 ```json
 {
   "mcpServers": {
-    "bridged": {
+    "fleetd": {
       "type": "http",
       "url": "http://127.0.0.1:8765/mcp"
     }
@@ -78,12 +78,12 @@ The entry to add, exactly:
 }
 ```
 
-If `.mcp.json` already exists, add only the `bridged` key and leave every other server untouched.
-If a `bridged` entry is already there with a different URL, **ask** rather than assuming yours is
+If `.mcp.json` already exists, add only the `fleetd` key and leave every other server untouched.
+If a `fleetd` entry is already there with a different URL, **ask** rather than assuming yours is
 right — a non-default port usually means a deliberate second daemon.
 
 > **If this plugin is installed, you can skip this step entirely.** The plugin ships its own
-> `.mcp.json`, so `bridged` is already mounted for any session with the plugin enabled. Write the
+> `.mcp.json`, so `fleetd` is already mounted for any session with the plugin enabled. Write the
 > project-level file only when the user wants the mount to work *without* the plugin — for
 > teammates who have not installed it, or for CI.
 
@@ -111,18 +111,18 @@ project already set.
   "$schema": "https://json.schemastore.org/claude-code-settings.json",
   "permissions": {
     "allow": [
-      "mcp__bridged__bridge_whoami",
-      "mcp__bridged__bridge_list",
-      "mcp__bridged__bridge_status",
-      "mcp__bridged__bridge_profiles",
-      "mcp__bridged__bridge_poll"
+      "mcp__fleetd__fleet_whoami",
+      "mcp__fleetd__fleet_list",
+      "mcp__fleetd__fleet_status",
+      "mcp__fleetd__fleet_profiles",
+      "mcp__fleetd__fleet_poll"
     ]
   }
 }
 ```
 
-Only the **read-only** bridge verbs are pre-allowed. `bridge_spawn`, `bridge_send`, and
-`bridge_stop` start processes, deliver work, and tear down terminals — those stay behind a prompt
+Only the **read-only** bridge verbs are pre-allowed. `fleet_spawn`, `fleet_send`, and
+`fleet_stop` start processes, deliver work, and tear down terminals — those stay behind a prompt
 on purpose. Do not "helpfully" add them.
 
 Never write `settings.local.json` on the user's behalf; that file is personal and usually
@@ -157,7 +157,7 @@ Write none of these into any file. Show the user the `export` lines to run thems
 Reconnect MCP if needed (`/mcp`), then confirm the tools are live and this session is the primary:
 
 ```
-bridge_whoami
+fleet_whoami
 ```
 
 - `{"role":"primary"}` — correct, you are done with this step.
@@ -169,7 +169,7 @@ bridge_whoami
 
   ```yaml
   primary:
-    terminal: term_xxxxxxxxxxxx     # the terminalId bridge_whoami just reported
+    terminal: term_xxxxxxxxxxxx     # the terminalId fleet_whoami just reported
   ```
 
   The daemon reads this **at boot**, so it needs a restart. Re-pin whenever the primary moves
@@ -178,9 +178,9 @@ bridge_whoami
 Then prove the fleet actually works, with a real spawn:
 
 ```
-bridge_profiles                       → the configured backends
-bridge_spawn{profile: "<one of them>"} → must reach state "ready"
-bridge_stop{paneId: "<from spawn>"}
+fleet_profiles                       → the configured backends
+fleet_spawn{profile: "<one of them>"} → must reach state "ready"
+fleet_stop{paneId: "<from spawn>"}
 ```
 
 **`/healthz` reporting `ok` is not evidence that spawning works.** It reports that the daemon can
@@ -216,7 +216,7 @@ State plainly:
 ```
 prereqs:   herdr <version> · bridged <protocol> · ccs <version> · codex <version|absent>
 written:   <files created or merged, or "none">
-role:      <bridge_whoami result — and the pin, if one was needed>
+role:      <fleet_whoami result — and the pin, if one was needed>
 spawn:     <real spawn result: profile, state reached, torn down>
 env:       <variables the USER still needs to export — names only, never values>
 skipped:   <anything not done, and why>

scripts/redeploy-bridged.sh

@@ -212,7 +212,7 @@ if [ -n "$OLD_PID" ] && [ "$ASSUME_YES" = 0 ]; then
   echo "   A restart drops every in-flight ticket and rendezvous. A member's report"
   echo "   is NOT recoverable once its ticket is gone."
   echo
-  echo "   Confirm with bridge_list that no members are live, and bridge_poll anything"
+  echo "   Confirm with fleet_list that no members are live, and fleet_poll anything"
   echo "   you still want, BEFORE continuing."
   echo
   read -r -p "   Fleet drained? type yes to restart: " reply
@@ -356,6 +356,6 @@ else
   ok "no ERROR lines since restart"
 fi
 echo
-echo "   Next: call bridge_whoami and confirm it still answers 'primary'. A lead whose tab label"
+echo "   Next: call fleet_whoami and confirm it still answers 'primary'. A lead whose tab label"
 echo "   no longer matches fleet.leaders.*.tab is demoted to worker and refuses orchestration."
 echo