Repository navigation
fix(doctor): teach plugin checks the post-A manifest contract - #2633
Conversation
Live-QA finding from the v5.260723.3 dogfood: two doctor checks still demanded the pre-Group-A manifest shape (mcpServers pointing to ./.mcp.json), so every post-A plugin generation warned 'MCP launcher unusable' and 'manifest does not declare ./.mcp.json' on perfectly healthy machines — the exact declaration Group A removed by design. - 'Codex Genie plugin' health is now installed+enabled+current; the plugin ships no MCP surface, so launcher usability is diagnostics, and the probe's stale manifest complaint (the healthy post-A shape) is filtered from it - 'Codex Genie MCP capability' inverts: declaring NO mcpServers is the pass state; a declared route is the warned regression (second Genie route risks cache-root routing); missing/corrupt/unproven arms unchanged The probe's own usability semantics are deliberately untouched: flipping 'usable' to true on post-A manifests would resurrect the plugin-route arms in route reconciliation/inspection (conflict + fallback removal). Retiring those arms wholesale is the tracked deeper follow-up.
|
Caution The consumer version of Gemini Code Assist on GitHub has been sunset. All code review activity has officially ceased. |
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Plus Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 878462cc97
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| const parsed = JSON.parse(readFileSync(manifest, 'utf8')) as unknown; | ||
| if (typeof parsed === 'object' && parsed !== null && !Array.isArray(parsed)) { | ||
| manifestState = | ||
| (parsed as Record<string, unknown>).mcpServers === undefined ? 'declares-none' : 'declares-route'; |
There was a problem hiding this comment.
Warn when the Codex manifest still advertises MCP
When an active post-A manifest omits mcpServers but still has interface.capabilities containing MCP, this branch classifies it as declares-none and the Codex Genie MCP capability check passes. The repo's post-A contract and release smoke checks reject both a mcpServers route and an advertised MCP capability, so genie doctor would miss a cached/installed plugin that still exposes the removed MCP surface in its metadata; include the capabilities field in this manifest-state check before returning a pass.
Useful? React with 👍 / 👎.
…s, IN_PROGRESS The H3 SessionStart line surfaced ledger drift: waves 1+2 shipped 2026-07-22 (PRs automagik-si#2625/automagik-si#2626/automagik-si#2628/automagik-si#2629, per-group SHIP reviews, validations reproduced) but their evidence and criteria were never recorded; status still read APPROVED with execution long underway. - status APPROVED -> IN_PROGRESS - 41 evidence-backed criteria ticked (5 -> 46/67): Groups A-D ACs + global criteria owned by merged groups; deliberately NOT ticked: the two live 'codex mcp get genie --json' operator proofs, Group F/G criteria, and all post-merge QA rows - appended the waves-1+2 evidence block, the seven live-QA fixes (automagik-si#2631-automagik-si#2634, automagik-si#2636), Group E's actual merge commit (4be6917), and the open automagik-si#2633-deferred follow-up (pre-A route-arm retirement) Adversarially verified pre-commit by the pm-ledger-verify workflow: 0 must-fix; its 4 advisory findings (defect count, untracked follow-up, unrecorded E merge) are incorporated above.
Live-QA finding from the v5.260723.3 dogfood (mac doctor)
Both warns are false alarms: they demand the pre-Group-A manifest shape —
mcpServers: ./.mcp.json— which Group A removed by design (Decisions 1/7: the marker-owned project route is the only Codex route). Every healthy post-A plugin generation triggers them forever.Fix (doctor-side only)
Deliberately NOT changed
The probe's
usable/activeManifestErrorsemantics are untouched: accepting the post-A manifest there would flipisUsableCodexPluginto true on healthy machines, resurrecting the plugin-route arms in route inspection/reconciliation (conflictfindings everywhere + fallback removal). Retiring those pre-A route arms wholesale is the tracked deeper follow-up and needs its own reviewed pass.Validation (self-run): doctor 106/0 (new regression test pins both manifest directions), PTY flow 9/0, doctor-observation 7/0; typecheck + lint clean.