CB-622 lead follow-up: the three changes no worker could make #136

2026-08-22T21:59:40+02:00

ltms commented

2026-08-22 21:59:40 +02:00

Closes the gaps left by CB-622 Units A/B/C (#135, #133, #132). Four files, three insertions, three deletions.

1. opencode.json — mount key bridged -> fleetd.
Unit C was briefed to do this and could not. opencode.json is in GitWorktrees.WORKTREE_HOSTILE_CONFIGS, so every worktree gets a neutral stub flagged --skip-worktree. The worker read the stub, found no mount key, and reported that accurately — a truthful conclusion that was wrong about the repo. The blind spot itself is CB-628 (#134); this is just the file.

2. e2e/bridge_ask_transcript.md keeps its name.
Unit B renamed it to fleet_ask_transcript.md (R100, content untouched). But the file is a dated record of a live run, and its own first line reads:

# Live bridge_ask — reverse rendezvous — 2026-07-16 16:30

That run really did call bridge_ask. Renaming the file made its name disagree with its content and its date, so the rename is reverted. Instead e2e/bridge_ask_test.py:213 now writes fleet_ask_transcript.md, so future runs get the new name while the old record stays intact. The harness header already said Live fleet_ask, so filename and content now agree on both sides.

Neither worker could resolve this alone: Unit B owned the .md, Unit C owned the .py, and the fix needs both. Both flagged it rather than guessing.

3. docs/MCP-Contract.md line 20.
The historical banner describes §6 and named bridge_ask; §6 now says fleet_ask. My scope instruction to leave the banner alone caused the drift, so this corrects it.

Verification

mvn clean install on the full integration of #135 + #132 + #133 + this: Tests run: 878, Failures: 0, Errors: 0, Skipped: 0, BUILD SUCCESS.

python3 -m py_compile e2e/bridge_ask_test.py passes. python3 -m json.tool opencode.json passes.

Still outstanding after this merge

Live MCP verification. Every test so far asserts on what the code builds, not on what an MCP client accepts. That is exactly how CB-618 shipped green and broke every role spawn. The daemon must be redeployed and fleet_whoami plus a deprecated bridge_whoami called for real.
CLAUDE.md and the wiki template, deliberately held: the canonical block tells every member to end its turn with bridge_reply, and renaming that before the running daemon answers fleet_* would silently lose the reports of any member spawned in the gap.

Closes the gaps left by CB-622 Units A/B/C (#135, #133, #132). Four files, three insertions, three deletions. **1. `opencode.json` — mount key `bridged` -> `fleetd`.** Unit C was briefed to do this and could not. `opencode.json` is in `GitWorktrees.WORKTREE_HOSTILE_CONFIGS`, so every worktree gets a neutral stub flagged `--skip-worktree`. The worker read the stub, found no mount key, and reported that accurately — a truthful conclusion that was wrong about the repo. The blind spot itself is CB-628 (#134); this is just the file. **2. `e2e/bridge_ask_transcript.md` keeps its name.** Unit B renamed it to `fleet_ask_transcript.md` (R100, content untouched). But the file is a dated record of a live run, and its own first line reads: ``` # Live bridge_ask — reverse rendezvous — 2026-07-16 16:30 ``` That run really did call `bridge_ask`. Renaming the file made its name disagree with its content and its date, so the rename is reverted. Instead `e2e/bridge_ask_test.py:213` now writes `fleet_ask_transcript.md`, so future runs get the new name while the old record stays intact. The harness header already said `Live fleet_ask`, so filename and content now agree on both sides. Neither worker could resolve this alone: Unit B owned the `.md`, Unit C owned the `.py`, and the fix needs both. Both flagged it rather than guessing. **3. `docs/MCP-Contract.md` line 20.** The historical banner describes §6 and named `bridge_ask`; §6 now says `fleet_ask`. My scope instruction to leave the banner alone caused the drift, so this corrects it. ## Verification `mvn clean install` on the full integration of #135 + #132 + #133 + this: **Tests run: 878, Failures: 0, Errors: 0, Skipped: 0, BUILD SUCCESS.** `python3 -m py_compile e2e/bridge_ask_test.py` passes. `python3 -m json.tool opencode.json` passes. ## Still outstanding after this merge - **Live MCP verification.** Every test so far asserts on what the code builds, not on what an MCP client accepts. That is exactly how CB-618 shipped green and broke every role spawn. The daemon must be redeployed and `fleet_whoami` plus a deprecated `bridge_whoami` called for real. - **`CLAUDE.md` and the wiki template**, deliberately held: the canonical block tells every member to end its turn with `bridge_reply`, and renaming that before the running daemon answers `fleet_*` would silently lose the reports of any member spawned in the gap.

ltms added 1 commit 2026-08-22 21:59:40 +02:00

CB-622 lead follow-up: the three changes no worker could make

CI / contract (pull_request) Successful in 1m10s

Details

CI / build (pull_request) Successful in 1m30s

Details

3aea4e1ec6

1. opencode.json — mount key bridged -> fleetd. Unit C could not do this:
   the file is neutralized by the worktree overlay, so every worker sees a
   stub and correctly reported it held no mount key. Tracked as CB-628.

2. e2e/bridge_ask_transcript.md keeps its name. It is a dated record of a
   run on 2026-07-16 that really did call bridge_ask, and its first line
   says so. The harness now writes fleet_ask_transcript.md for new runs;
   its header already says 'Live fleet_ask', so the two now agree.

3. docs/MCP-Contract.md line 20 — the historical banner describes section
   6, and section 6 now says fleet_ask.

ltms merged commit f4570ff274 into main

2026-08-22 21:59:46 +02:00

Sign in to join this conversation.

1 Participants

Notifications

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: fleet/fleetd#136