| name | gh-issue-sequencer |
| description | Maintain a clear next-work queue in GitHub Project #5 by normalizing issue status, priority, and execution order; route genuine priority tradeoffs to issue-audit. |
| category | github-issue |
| kind | atomic |
| phase | context |
| requires_write | true |
| requires_slurm | false |
| requires_benchmark_artifacts | false |
| delegates_to | [] |
| output_schema | skill_run_summary.v1 |
GH Issue Sequencer
Purpose
Maintain a clean Project #5 execution queue by separating clarification cleanup, readiness routing,
and one-pass sequencing metadata writes. Project #5 priority is advisory; current maintainer
direction and fresh evidence can override queue order. If ordering depends on a real value
tradeoff, route to issue-audit for one focused priority question rather than blocking autonomous
selection broadly.
Workflow
- Prepare:
- read
docs/project_prioritization.md and docs/context/issue_713_batch_first_issue_workflow.md,
- resolve project/field IDs once for the session,
- check
gh api rate_limit when batch size is large.
- Inspect queue:
- list Project #5 items and issue metadata,
- use REST for issue fields when GraphQL is constrained.
- Resolve blockers first:
- route ambiguous issues to
gh-issue-clarifier,
- route implausible priorities to
gh-issue-priority-assessor,
- route genuine value tradeoffs to
issue-audit priority discussion,
- keep
decision-required, blocked, duplicate, wontfix out of execution-ready ordering.
- Normalize issue status:
In progress for the active item,
Ready for actionable and unblocked work,
Tracked for valid but deferred work,
Done for merged/closed issues.
- Order and apply:
- sort by explicit maintainer direction first, then higher priority, lower uncertainty, unlock
factor, and oldest issue number,
- continue autonomous ordering when the top candidate is clearly actionable,
- use priority discussion only when two or more plausible next issues depend on a real value
tradeoff that repository evidence and Project #5 fields cannot resolve,
- apply status/priority/duration edits in one write pass,
- run score sync once at batch end if inputs changed.
- If writes fail (rate limits or auth), stop writes and capture exact pending mutation details.
Priority Discussion Mode
Use this mode sparingly. It exists for queue-ordering tradeoffs, not for routine approval of every
autonomous issue selection.
Ask one focused priority question only when all are true:
- at least two unblocked issues are implementable now,
- their Project #5 score or labels do not clearly decide the order,
- the choice changes real maintainer value, risk, or unblock impact,
- the answer cannot be inferred from recent issue comments, PR state, or canonical docs.
Frame the question around concrete choices, for example:
Should the next PR prioritize reducing CI minutes (#A) or improving benchmark provenance (#B)?
After the maintainer answers, record the decision in the relevant issue as a short comment or
Maintainer priority note body entry. If the answer changes Project #5 priority fields, batch that
write with the normal Project #5 metadata pass and run score sync once at the end. If the answer is
only an ordering preference, do not invent new priority-score inputs.
Guardrails
- Use MCP/interactive tools when available; use
gh for deterministic sequencing mutations.
- Do not interleave issue-body edits, project routing, and score sync issue-by-issue for multi-issue batches.
- Do not ask a priority question when the next issue is already clear, when a blocker/clarification
question is really needed instead, or when the tradeoff is only agent convenience.
- Use follow-up handoffs rather than retry loops when quotas are temporarily exhausted.
Output
- Ordered issue queue with one-line rationale per item.
- Priority discussion question asked, answer recorded, or reason it was unnecessary.
- Status and priority changes applied.
- Unresolved blockers and next candidate issue.
- Whether final score sync completed.
When to use
Use this skill for the scope named in its frontmatter description and registry metadata.