Support epics and human checkpoints - #33
Conversation
![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: 40f6f79b4d
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".
| or ( | ||
| isinstance(ticket["labels"], list) | ||
| and args.human_work_label in ticket["labels"] | ||
| ) |
There was a problem hiding this comment.
When a Backlog human-work issue is assigned to the authenticated user and any normalized field raises InputError (for example, a missing projectPosition), this new branch identifies it as human work, but the exception path subsequently classifies every current-user Backlog assignment as blockedPlanningClaims. The scheduler treats that collection as unresolved runner work and prevents triage, even though valid assigned human work is deliberately not a runner claim. Preserve the human-work role during error classification so these entries remain frontier/excluded items, and cover the malformed assigned-human counterexample.
AGENTS.md reference: AGENTS.md:L18-L18
Useful? React with 👍 / 👎.
| Finish successfully only when the ordinary drain finish gate passes and no | ||
| human action remains. Return partial drain only for a failed, ambiguous, | ||
| blocked, or timed-out claimed lane. Never call a human frontier a failure or |
There was a problem hiding this comment.
When the triage provider cannot complete an eligible item, drain-scheduler.md explicitly requires a partial-drain report, but this finish gate limits partial drain to failures in a claimed lane; the triage lane has no claim. This leaves conflicting outcome instructions, particularly when human actions also remain, so include triage-provider failures in this partial-drain condition or explicitly defer to the scheduler's authoritative finish gate.
AGENTS.md reference: AGENTS.md:L16-L16
Useful? React with 👍 / 👎.
|
| Filename | Overview |
|---|---|
| skills/run-github-project/scripts/rank_tickets.py | Adds validated Backlog role classification and separate outputs for epics, human actions, triage candidates, and blocked frontier work. |
| skills/run-github-project/scripts/test_rank_tickets.py | Adds deterministic coverage for epic readiness, human gates, dependency parking, conflicting roles, and assigned human work. |
| skills/run-github-project/references/human-frontier.md | Defines classification, reconciliation, reporting, resumption, and finish semantics for epics and human checkpoints. |
| skills/run-github-project/SKILL.md | Integrates the new frontier into startup validation, scheduling, authority checks, reconciliation, and reporting. |
| skills/run-github-project/references/drain-scheduler.md | Places epic reconciliation in the controller lane and adds waiting-for-human behavior to the drain finish gate. |
Reviews (2): Last reviewed commit: "Address human frontier review feedback" | Re-trigger Greptile
1 participant
Summary
run-github-projectWhy
Coordination parents and human gates previously appeared only as blocked or triage work, preventing the drain from expressing actionable manual checkpoints while independent leaf work continued.
Validation
python3 skills/run-github-project/scripts/test_rank_tickets.py(65 tests)npm run lintpython3 /Users/chris/.codex/skills/.system/skill-creator/scripts/quick_validate.py skills/run-github-projectgit diff --check origin/main...HEADreview-and-simplify-changes,code-review, andponytail-review