Skip to content

KamiJeong/tech-bridge-book / .agents/skills/speckit-gh-issue-runner

speckit-gh-issue-runner

Verified

Find up to three open GitHub issues marked for $speckit-auto, inspect them, and run the existing speckit-auto workflow sequentially with issue labels and comments.

Spec Kit GitHub Issue Runner

Use this skill when the user asks to check GitHub issues, take issues from GitHub, process issues labeled for the Spec Kit auto workflow, or run a maximum of three issues through speckit-auto.

This skill only owns issue discovery and queue handling. After an issue is selected, follow .agents/skills/speckit-auto/SKILL.md for the actual workflow and .agents/skills/speckit-token-observability/SKILL.md for required token analysis.

Limits

  • Default maximum: 3 issues.
  • Never process more than 3 issues in one request, even if the user asks for more.
  • Process issues sequentially unless the user explicitly asks only for a dry-run queue summary.
  • Prefer gated mode unless the issue explicitly requests auto-implement or auto-pr.
  • Do not auto-commit or create a PR unless the issue explicitly grants auto-commit or auto-pr permission.

Discovery

From the repository root, list candidate issues:

.agents/skills/speckit-gh-issue-runner/scripts/list-speckit-auto-issues.sh

The script returns up to three open issues labeled trigger:speckit-auto. Discovery MUST fetch status:queued candidates first. If fewer than three queued issues are available, fetch a wider candidate set before applying the three-issue cap, then filter and rank by workflow status so completed or blocked issues do not hide runnable work.

If the script fails because gh is unavailable or unauthenticated, use the GitHub app tools if available. If neither route works, stop and report the blocker.

Selection Rules

When the user gives explicit issue numbers, use those numbers after enforcing the max-3 limit.

When the user asks to "check issues" without numbers:

  1. List up to 3 candidates.
  2. Exclude issues already labeled status:in-progress, status:blocked, status:waiting-review, status:done, or status:failed.
  3. Prefer issues labeled status:queued by querying queued issues first.
  4. Otherwise prefer oldest updated issue first, unless priority labels indicate priority:P1, then priority:P2, then priority:P3.
  5. If selection is still ambiguous, present the candidate numbers and ask the user which ones to run.

Intake For Each Issue

For each selected issue:

  1. Fetch issue title, body, labels, comments, checklists, and linked sub-issues.
  2. Confirm the issue has the trigger:speckit-auto and agent:codex labels, or $speckit-auto appears in the title, body, or comments.
  3. Determine mode from issue labels or issue template text:
    • mode:gated, (gated), "plan first", or "wait for approval" -> gated
    • mode:auto-implement, (auto-implement), or "safe, low-risk changes" -> auto-implement
    • mode:auto-pr, (auto-pr), or "one draft pull request" -> auto-pr
    • Korean choices map as follows: "먼저 계획" or "승인한 뒤" -> gated, "안전하고 작은 변경" -> auto-implement, and "검토용 요청서 하나" -> auto-pr
  4. If no mode is present, use gated.
  5. Add or preserve trigger:speckit-auto, agent:codex, exactly one mode:* label, and status:in-progress.
  6. Add a short issue comment before starting intake.

Running The Workflow

Run the full issue through speckit-auto using the issue body as the feature request. Follow all stop conditions, token observability requirements, and phase label/comment rules from speckit-auto.

If one issue stops for clarification, blocking gates, failed validation, or review, update that issue with the correct status label and continue to the next selected issue only when the stop is isolated to that issue.

If repository state becomes unsafe for the remaining issues, stop the batch and report which issue caused the stop and which selected issues remain unprocessed.

Final Response

Report:

  • Issue numbers processed
  • Final status label for each issue
  • Mode used for each issue
  • Token report path when created
  • Any issue numbers skipped and why
planning-and-task-breakdownaddyosmani/agent-skillsBreaks work into ordered tasks. Use when you have a spec or clear requirements and need to break work into implementable tasks. Use when a task feels too large to start, when you need to estimate scope, or when parallel work is possible.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
screenpipethedotmack/claude-memSearch screen and audio history recorded by Screenpipe and optionally save selected context into claude-mem. Use when the user asks to recall desktop activity, a meeting, or something seen in another app through Screenpipe.Stars: 118kUpdated Claude CodeCodexVerifiedInstall
contact-widgetnexu-io/open-designSelf-contained floating chat widget with welcome screen, social links, meeting button, and message input. Single HTML file, zero dependencies.Stars: 108kUpdated Claude CodeCodexVerifiedInstall
meeting-notesnexu-io/open-designMeeting notes page — title bar with attendees, agenda checklist, decisions block, action items table with owners + dates, and a "next meeting" footer. Use when the brief mentions "meeting notes", "minutes", "1:1 notes", "all-hands recap", or "会议纪要".Stars: 108kUpdated Claude CodeCodexVerifiedInstall