菜单
When a dispatched run produces text or findings rather than a PR -- use this to pick the right output destination (stdout, CLAUDE.md entry, issue comment, findings file), apply the spawn-issue handback pattern, and hold the synchronous discipline for non-PR runs.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use when deciding whether to dispatch a runner or a worker for a subtask. Describes the decision boundary, the failure mode when runner is used where worker is intended, and the safe composition pattern.
When a repo has no GitHub remote (or is not yet a git repo) -- use this to route survey, sweep, and dispatch off the GitHub-centric path. Detect with `git remote` empty; treat blank GitHub columns as expected, not an error; route work to commits + CLAUDE.md instead of issues + PRs; degrade a maintenance sweep to the local-only axes.
When the directive is "status check," "what's new across my projects," or "sweep maintenance" -- run a per-project, read-mostly, non-PR sweep. For each project under the owner-prefixed layout, gather stars/forks, open issues/PRs split into mine vs community vs bots, package downloads, release-due (unreleased conventional commits), a test run, a stale-CLAUDE.md flag, and branch-cleanup candidates. Emit a fresh per-project report every sweep; NEVER cache status back into a workspace/manager CLAUDE.md.
Before making any non-trivial code, docs, or config change -- create a feature branch first, then open a PR. Never commit directly to main. Apply by default before any edit; this is the baseline branch discipline that all other git skills build on.
When the issue or PR queue has open unlabeled items -- run a read-only triage pass that labels each one by component, category, priority, and size, flags duplicates, closes noise, and reports the p1 queue before runners are dispatched.
When writing any issue, PR, or commit message, use this for the authoring standard. Covers structure (headings, lists, tables, fenced blocks, the bug block, the before/after) and voice (factual, no emdashes, no editorializing labels, conventional-commit prefixes, no commit trailers).
| name | non-pr-output-conventions |
| description | When a dispatched run produces text or findings rather than a PR -- use this to pick the right output destination (stdout, CLAUDE.md entry, issue comment, findings file), apply the spawn-issue handback pattern, and hold the synchronous discipline for non-PR runs. |
Not every dispatch ends with a merged PR. Research, audits, Q&A, diagnosis, and status reports produce text or structured findings. This skill defines where that output goes and how the lifecycle differs from the PR-centric path.
| who needs it | duration | destination |
|---|---|---|
| Human reading now | ephemeral | stdout / return text |
| Human + future agents, this project | durable, project-scoped | CLAUDE.md entry (brainstorm-sketches or decisions log) |
| Visible to others, indexed by GitHub | durable, public | issue comment (gh issue comment N --body "...") |
| Cross-session, shareable artifact | durable, file | markdown file in repo or /tmp/<task-id>-findings.md |
| Spawns further work | structured input | issue body / issue comment as findings list |
Pick the most durable destination that the consumer actually needs. A quick Q&A answer belongs in stdout. An audit summary that a runner will act on belongs in an issue comment or findings file. Don't write to CLAUDE.md for answers a human won't search for later.
Just return it in the response. No file, no issue, no branch. Right for ephemeral answers: "explain how --fresh works," "what shipped this week?"
Use the existing brainstorm-sketches or decisions-log pattern from
orchestration-prompt-template. Append to the relevant section;
no commit needed (CLAUDE.md is typically untracked).
gh issue comment N --body "$(cat /tmp/<task-id>-findings.md)"
Right for audit findings, diagnosis results, research summaries that belong to a tracking issue. Visible to all; searchable.
Write to /tmp/<task-id>-findings.md. Return the path in the
response so the dispatcher can reference it in a follow-on prompt.
Use when a subsequent runner dispatch will consume the findings.
When a non-PR run surfaces an actionable finding (e.g. an audit finds an exploitable bug), file a new issue and return its number:
gh issue create \
--title "fix: <short description of finding>" \
--body "$(cat /tmp/<task-id>-findings.md)" \
--label "fix,p1"
If priority warrants immediate action, the dispatcher can dispatch a runner on the new issue directly. The auditor does NOT dispatch the runner itself -- that decision belongs to the dispatcher.
Non-PR runs carry the same hold-open contract as PR runs. The
dispatched session must complete and deliver to the durable
destination before returning. A non-PR session that fires
run_in_background=true and returns has orphaned the output --
same failure mode as an unmerged PR.
No branch, no empty commit, no CI watch. These are not needed and
add overhead. Everything else from runner-synchronous-lifecycle
applies.
| shape | model | effort | rationale |
|---|---|---|---|
| Q&A / explainer | haiku | low | Shallow read, short answer |
| Status / report | haiku | low | Structured retrieval, brief synthesis |
| Research | sonnet | high | Multi-source reasoning, cited output |
| Audit | sonnet | high | Deep scan, structured findings |
| Diagnosis | sonnet | medium | Focused analysis, RCA output |
These are defaults. The dispatcher overrides per the specific task.
runner-synchronous-lifecycle
-- the hold-open contract this skill extends to non-PR runsorchestration-patterns
-- researcher and auditor execution shapesdispatch-options
-- model: and effort: hints for non-PR dispatches