agentleFS
salvo8519salvo8519/fairmark/.agents/skills/run-sprint/SKILL.md

run-sprint

Skill1 starsChanged 2 days agoApache-2.0View on GitHub

Orchestrate an approved sprint — delegate every issue to the specialised agents lane by lane (locally) or dispatch issues to Claude in GitHub Actions (remotely), ending each issue as a pull request. Use when the maintainer asks to run or continue a sprint.

  • Commits and pushes

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 1 is distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read the issue (gh issue view N) and its spec.

    line 17
  2. Very specific:

    No approved spec for method work: delegate to spec-writer, then stop this lane and ask the maintainer to approve.

    line 18
  3. Distinctive:

    Keep a running table: issue, branch, PR, status, blockers.

    line 37
  4. Very specific:

    when a lane needs generated artifacts, it becomes the generate lane of the sprint, the only one (sprints/README.md).

    line 41
  5. Very specific:

    Do not change status to done: that is the maintainer's call.

    line 55

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

The main session is the orchestrator: subagents cannot start other subagents, so you delegate and keep the overview. Ask the maintainer which mode to use if not stated.

Preconditions

  • sprints/SNN.yaml exists with status: active and passes uv run python scripts/sprint.py validate.
  • main is green.

Local mode (subagents, parallel lanes)

For each lane, in parallel where the agents tool supports isolated worktrees, otherwise one lane after the other. For each issue of the lane, in order:

  1. Read the issue (gh issue view N) and its spec. Create branch sprint/SNN/N-<slug>.
  2. No approved spec for method work: delegate to spec-writer, then stop this lane and ask the maintainer to approve.
  3. test-engineer writes failing tests for the issue's acceptance criteria.
  4. core-engineer (Rust core) or docs-writer (documentation-only issues) implements.
  5. Review by tier ("Effort proportional to risk" in AGENTS.md): none for small pull requests, one code-reviewer pass for substantial ones, plus numerical-auditor for numerical changes and security-reviewer when dependencies or workflows changed. Fix every finding; a second pass only for a blocking logic defect, then escalate to the maintainer. Run at most three lanes at a time.
  6. uv run just check green, docs in sync (the Stop hook enforces it).
  7. Push and open the PR: gh pr create --title "<type>(<scope>): ... [SPEC-NNNN]" --body "Closes #N ...". Never merge. Checkpoint state at every step: after each numbered step of this loop, run uv run python scripts/sprint.py state sprints/SNN.yaml N --step <step> [--pr <n> | --blocker "<why>" | --clear-blockers] (the running table persisted, sprints/README.md, Sprint state). At every task switch append one line to .agents/journal/<date>-sprint-SNN-wip.md: issue, step, pull request, blockers, last commit of the lane. Both are local recovery state, git-ignored: a crash costs minutes, not a session of context. Keep a running table: issue, branch, PR, status, blockers. Each lane works in its own worktree (one worktree per session, AGENTS.md):
  • a lane other than the generate lane never runs uv run just generate;
  • when a lane needs generated artifacts, it becomes the generate lane of the sprint, the only one (sprints/README.md).

Remote mode (Claude in GitHub Actions)

For each issue: gh issue comment N --body "@claude Implement this issue following .agents/skills/implement-spec/SKILL.md and AGENTS.md. Work test-first, run uv run just check, and open a pull request that closes this issue." Issues of different lanes run in parallel on separate runners. Then track the resulting PRs with scripts/sprint.py status.

End of run

Propose the merge order to the maintainer (sprints/README.md, Generated artifacts and the merge order): pull requests carrying regenerated artifacts merge strictly one at a time; after each of them merges, rebase the next on main and run uv run just generate in it again. Report per lane: PRs opened, issues blocked and why, what the maintainer must review or decide. Do not change status to done: that is the maintainer's call. Propose the removal of the worktree of every merged pull request: the Retirement candidates of uv run just workspace-report, never deleted without the maintainer's approval.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.