fix(linear): rename the agent:* label family to ai:* in AGENTS.md - #18
Conversation
Closes PRD-756. Every routine spec and agent-facing doc instructed writers to apply `agent:routine` / `agent:codex` / `agent:claude`. No `agent` label group exists in Linear. The live group is `ai`, with exactly those three children. `save_issue` rejects the entire label array when any one entry fails to resolve, so an agent following the docs literally got a hard write failure and filed nothing, which is how health-watch hit this while filing PRD-755. Two traps this rename had to avoid: - `package:agent` is a real, unrelated label (the agent runtime package) and is untouched. This was not a find-and-replace of the word "agent". - `agent:copilot` appeared in the operating model as an example, and has no `ai:` equivalent at all: the live group is only routine, codex, and claude. Dropped rather than translated. Also documents the second failure mode from the same issue: the `group:child` notation used throughout these docs is display shorthand, not accepted API input. `save_issue` resolves by bare child name or ID, so `["protocol:green-goods", "activity:qa"]` fails where `["green-goods", "qa"]` succeeds. PRD-756`s own label array reads `[routine, qa, docs, green-goods]`, which is the proof. Stated at the canonical definition and at both routine-facing indexes, since that is where writers look. One further correction found while reading: the guild routines README said to "optionally add `ai:claude` or `ai:codex`" on top of `ai:routine`. Linear enforces one child per grouped family, so that write would be rejected for the same reason. It now says they replace rather than accompany. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
|
Warning Review limit reached
Next review available in: 56 minutes Enable usage-based reviews in Billing to review now. Otherwise, wait until the next included review is available. How can I continue?After more reviews become available, a review can be triggered using the To avoid repeated limits, reduce automatic review volume by pausing incremental auto-reviews earlier, using label-based review opt-in, excluding WIP or generated PR titles, or requesting reviews manually when the PR is ready. If your team needs uninterrupted high-volume reviews, an organization admin can enable usage-based reviews. How do review limits work?CodeRabbit enforces per-developer PR review limits for each organization. Most developers receive the normal plan review availability. For paid Pro and Pro+ PR reviews, CodeRabbit uses adaptive limits for sustained high-volume activity. When a developer's recent PR review activity reaches the 95th percentile or higher among CodeRabbit users, additional reviews become available more gradually as earlier reviews age out of the rolling window. Please refer docs for additional details. ✨ 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 |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: b714765885
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".
| ## Linear-Spawned Issue Contract | ||
|
|
||
| Linear (workspace `greenpill-dev-guild`; teams Product `PRD` / Research `RESR`) is the durable backlog — never open GitHub Issues for backlog work. Codex work is labeled `agent:codex` (distinct from `agent:claude` interactive and `agent:routine` cron writes). The Linear MCP is wired into the Codex environment for read/query and state transitions. | ||
| Linear (workspace `greenpill-dev-guild`; teams Product `PRD` / Research `RESR`) is the durable backlog — never open GitHub Issues for backlog work. Codex work is labeled `ai:codex` (distinct from `ai:claude` interactive and `ai:routine` cron writes). The Linear MCP is wired into the Codex environment for read/query and state transitions. |
There was a problem hiding this comment.
When an agent follows this instruction while creating a Linear issue through the stated MCP, ai:codex (and likewise ai:claude or ai:routine) is still rejected because save_issue accepts a bare child name or ID rather than group:child notation—the second failure mode described in this commit. The rename therefore leaves the hard-write failure intact; identify these values as display shorthand and tell MCP callers to pass codex, claude, or routine.
Useful? React with 👍 / 👎.
1 participant
Closes PRD-756.
Every routine spec and agent-facing doc told writers to apply
agent:routine/agent:codex/agent:claude. Noagentlabel group exists in Linear. The live group isai, with exactly those three children. Becausesave_issuerejects the entire label array when any one entry fails to resolve, an agent following the docs literally got a hard write failure and filed nothing. That is howhealth-watchhit this while filing PRD-755.Verified against live Linear before changing anything: the
aigroup holdsroutine(90260e40…),codex(50724377…),claude(33e73a3a…), and there is noagentgroup.Two traps this rename had to avoid
package:agentis a real, unrelated label (the agent runtime package) and is untouched. This was not a find-and-replace of the word "agent"; the substitutions were scoped to the four label tokens.agent:copilotappeared in the operating model as an example and has noai:equivalent at all. The live group is only routine/codex/claude, so it was dropped rather than translated.The second finding, also from PRD-756
The
group:childnotation these docs use throughout is display shorthand, not accepted API input.save_issueresolves labels by bare child name or ID, so["protocol:green-goods", "activity:qa"]fails where["green-goods", "qa"]succeeds. PRD-756's own label array reads[routine, qa, docs, green-goods], which is the proof. This is now stated at the canonical definition and at both routine-facing indexes, since that is where writers actually look.One further correction found while reading
The guild routines README said to "optionally add
ai:claudeorai:codex" on top ofai:routine. Linear enforces one child per grouped family, so that write would be rejected for exactly the same reason as the original bug. It now says they replace rather than accompany.Not in this PR
green-goods/.plans/active/commitment-pooling/**also carriesagent:*references. Left alone deliberately: the audit reports andstatus.jsonevent log are dated historical records that should not be rewritten, and the lane data is live input toplan-hub.mjs linear-syncfor a feature another agent is actively working. It needs its own scoped pass, with theplan-hub.test.mjsfixtures moving in lockstep.🤖 Generated with Claude Code