agentleFS
runxhqrunxhq/runx/skills/issue-to-pr/SKILL.md

issue-to-pr

Skill97 starsChanged 33 days agoApache-2.0View on GitHub

Implement one bounded repository issue with normal host tools, prove the tested change through one scafld finalize wall, and optionally publish the exact pull request through scoped provider authority with readback.

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 2 are distinctive, and 5 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 2

    Distinctive

    This author's own choices. Worth a look.

  • 5

    Very specific

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

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

    Read through the GitHub instance selected here: the signed-in gh session (runx connector use github --session) or a stored token.

    line 17
  2. Distinctive:

    Run one preflight before work: issue identity, repository permission, branch state, required tools, requested outcome, and publication …

    line 19
  3. Very specific:

    Call scafld finalize exactly once after the change and tests are ready.

    line 23
  4. Distinctive:

    If PR publication is not authorized, return the tested/finalized result and stop.

    line 27
  5. Very specific:

    If publication is authorized, pass the exact host_result to publish.

    line 29
  6. Very specific:

    Do not claim completed without a real edit/test outcome and one successful finalize result.

    line 84
  7. Very specific:

    No GitHub instance selected: return the exact runx connector use github --session or runx credential set github handoff.

    line 92

“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “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

Issue to PR

Turn one issue into a tested change and, when authorized, a verified pull request. Use the repository's normal development workflow. Runx owns evidence continuity and the consequential publication boundary; it does not replace the host agent's editor, shell, Git, tests, or repository judgment.

Direct operator flow

  1. Name the repository as an exact owner/name. Never infer the target from an available grant.
  2. Read through the GitHub instance selected here: the signed-in gh session (runx connector use github --session) or a stored token.
  3. Run one preflight before work: issue identity, repository permission, branch state, required tools, requested outcome, and publication authority.
  4. Investigate, edit, and test with the host agent's ordinary tools. Do not manufacture Runx answer files between normal coding steps.
  5. Call scafld finalize exactly once after the change and tests are ready. Preserve its workspace-scoped receipt path, exact target commit, and contract digest in host_result. Runx verifies that receipt with scafld; host-authored status strings are not proof.
  6. If PR publication is not authorized, return the tested/finalized result and stop. Do not silently downgrade the work to a plan.
  7. If publication is authorized, pass the exact host_result to publish. Runx pushes the exact commit to the requested branch (github/branch.push, never forced, read back from GitHub), reuses the one open pull request for that branch or opens one (github/pullrequest.open, read back), and completes only when the pull request's head is the tested commit. Notifications, feeds, issue comments, and documentation sync are optional downstream skills.

Reuse in chains

  • from-evidence accepts runx.issue_to_pr.issue_evidence.v1 and skips GitHub discovery and issue read.
  • resume accepts both prior issue evidence and a completed runx.issue_to_pr.host_result.v1; it does not repeat host work or finalize.
  • verify verifies the signed scafld receipt against the exact commit and contract without publishing anything.
  • finalize-local verifies the completed host result and returns it without a remote mutation.
  • publish verifies the completed host result, publishes its exact Git ref, and creates or recovers the authorized PR with readback.
  • Preserve the same idempotency key across pause, retry, and resume. An uncertain PR creation never gets a new key.

Host work contract

The issue-to-pr-host-work act performs normal repository work and returns:

host_result:
  schema: runx.issue_to_pr.host_result.v1
  status: completed | blocked | failed
  repository: owner/name
  issue_number: string
  repo_root: workspace-relative-or-absolute-path
  branch: string
  commit: full-git-object-id
  files: [relative/path]
  tests:
    - command: string
      status: passed | failed
      evidence: string
  finalization:
    receipt_path: workspace-relative-path
    contract_digest: sha256:...
  publication:
    decision: hold | ready
    title: string
    body: string
    head: string
    base: string
    draft: boolean
    idempotency_key: string
  errors: [string]

Do not claim completed without a real edit/test outcome and one successful finalize result. Do not claim published from a branch push, API acknowledgement, or draft packet; only independent PR readback closes that state.

Stop conditions

  • Wrong or ambiguous repository: stop with one target-resolution blocker.
  • No GitHub instance selected: return the exact runx connector use github --session or runx credential set github handoff; do not cycle through unrelated skills.
  • Dirty or conflicting branch state: stop before mutation and preserve the evidence already gathered.
  • Failed tests or failed/stale finalize: return blocked or failed, never succeeded.
  • Missing publication authority: retain the tested finalized change locally and stop before PR creation.
  • Provider failure after admission: preserve the grant reference, idempotency key, finalization evidence, and mutation recovery state. Resume at publication; do not repeat issue discovery, coding, tests, or finalize.

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.