agentleFS
phatnguyen03022001phatnguyen03022001/agent-skills/github-workflow/SKILL.md

github-workflow

Skill2 starsChanged 3 months agoApache-2.0View on GitHub

Use when target-authoritative Git/GitHub branch topology, commits, pushes, promotion, or GitHub Actions must be governed without inventing branches.

23 other projects use a version of this.

Adopted by

Most recently taken up by

Used with

Projects with it also use marketing-psychology, copywriting and 4 more.

Other files people use alongside this one.

Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
  • SentrySkill

    marketing-psychology

    When the user wants to apply psychological principles, mental models, or behavioral science to marketing. Also use when the user mentions 'psychology,' 'mental models,' 'cognitive bias,' 'persuasion,' 'behavioral science,' 'why people buy,' 'decision-making,' or 'consumer behavior.' This skill provides 70+ mental models organized for marketing application.

    Also in 3 projects that use this

  • SupabaseSkill

    copywriting

    Write or audit UI copy (buttons, labels, empty states, error messages, tooltips, form text) anywhere in the monorepo. Load it before shipping or reviewing any user-facing text — including when copy is incidental to the task, like a new feature that adds buttons, toasts, dialogs, or validation messages.

    Also in 3 projects that use this

  • Microsoft AzureSkill

    git-workflow

    Squad branching model: dev-first workflow with insiders preview channel

    Also in 3 projects that use this

  • SentrySkill

    skill-creator

    Create new agent skills following the Agent Skills specification. Use when asked to "create a skill", "add a new skill", "write a skill", "make a skill", "build a skill", or scaffold a new skill with SKILL.md. Guides through requirements, writing, registration, and verification.

    Also in 4 projects that use this

  • GoogleSkill

    code-review

    Adversarial systems, concurrency, memory-model, and architecture code review protocol. Use this skill whenever asked to review code, PRs, diffs, or branches in Filament.

    Also in 3 projects that use this

  • BlockSkill

    frontend-design

    Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.

    Also in 3 projects that use this

The file

We split this skill into the rules it gives an agent

3 rules. 0 are common, 0 are distinctive, and 3 are very specific to this project.

  • 0

    Common

    Most of the 16 edited versions we read keep these.

  • 0

    Distinctive

    This author's own choices. Worth a look.

  • 3

    Very specific

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

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

    Never infer or create staging merely because DEV_STAGING_MAIN is supported.

    line 14
  2. Very specific:

    Do not auto-promote after push, review, CI, or verifier success.

    line 55
  3. Very specific:

    After exact promotion, derive PROMOTED_NOT_RELEASED until all separately authorized release actions are complete.

    line 59

“Common” is measured: kept by at least 6 in 10 of the 16 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

GitHub Workflow

Repository-specific branch policy and exact live refs outrank generic defaults. This existing workflow owner supports three target-authoritative topologies:

  • MAIN_ONLY: one working/stable main-style branch when target authority says so.
  • DEV_MAIN: mutable dev integration plus stable main; this preserves backward compatibility for existing dev/main repositories.
  • DEV_STAGING_MAIN: dev → staging → main only when explicitly activated by target authority.

Never infer or create staging merely because DEV_STAGING_MAIN is supported. Never create any branch from a generic topology default; branch creation always needs explicit target/task authority. Branch names and roles come from target authority, not this skill.

Before writes

Refresh exact repository, target branch, remote HEAD, expected base, repository-specific branch policy, write authority, workflow triggers, and the semantic capabilities required for the current phase when that mutable evidence is required by the pending consequence. Treat authorized remote Git state as canonical repository truth; local state is an execution copy. A fetch is an observation mutation that may update LOCAL_MUTABLE remote-tracking refs; a local commit is an authorized target-ref mutation when task authority permits it; an authorized push/promotion/report-or-review publication is a canonical publication mutation. Immediately before canonical publication, freshly resolve the REMOTE_MUTABLE refs/topology that gate the write, and after publication freshly prove the resulting remote identity. If the remote branch moved, that remote drift invalidates stale authority and fails closed. If a current-phase capability is unavailable, block before mutation. Do not require a later-phase capability to finish an earlier authorized phase.

A local clean/behind copy may be synchronized when authorized; local ahead or local dirty state is divergence or unknown work and must not be auto-pushed, reset, deleted, or silently adopted as authority.

task.git_authority governs Executor Git mutations. create_branch, commit, push, and promotion are independent. When create_branch: false, do not invoke branch creation even for tests or temporary work. Commit does not imply push, force push, stable-branch mutation, or promotion.

release_authority is separate again: create_version_tag, mutate_repository_metadata, and publish_release are independent permissions and are never inferred from commit, push, or promotion.

Semantic capability preflight

Use phase-specific semantic requirements rather than equating authority with whichever API happens to exist. Route from required capability and evidence: resolve currently available candidates, reject candidates lacking current authority or sufficient evidence, then choose the lowest sufficient expected cost/resource burden. When materially equivalent, prefer fewer context transfers and lower consequence. Cheaper/free never permits weaker correctness, exact identity, acceptance evidence, safety, or required native/remote verification.

Availability, quota, rate limits, and paid capacity are runtime evidence. Installation, provider identity, historical availability, or an earlier preflight does not prove current availability after material environment/quota change. If the selected surface becomes unavailable or quota-limited, fallback may use only another currently available candidate that is already authorized and still sufficient for the same required capability/evidence; otherwise use an explicitly permitted degraded mode or fail closed with the existing capability/blocking semantics. Do not create provider/account rotation, quota-evasion, credential-broker, or persistent availability registries.

  • EXECUTION: required repository content mutation and test/verification execution surfaces, including native verification when the task declares it mandatory for the current execution;
  • REVIEW: exact commit/report resolution needed by the current governing Architect;
  • VERIFICATION: designated authoritative exact-SHA verifier access;
  • PROMOTION: non-force update of the authorized stable target ref to the exact candidate;
  • RELEASE: only tag, repository-metadata, and release-publication capabilities for actions explicitly authorized.

Unavailable capability blocks only the current phase. For example, a successful promotion followed by missing release publication capability yields valid PROMOTED_NOT_RELEASED rather than invalidating promotion.

Promotion candidate lineage

Let R = reviewed_report.commit for the exact report accepted by Architect.

A valid promotion_candidate_head is only:

  1. R; or
  2. the single-parent direct child of R, whose only parent is R, with that single commit containing only the expected Architect-owned review artifact.

A merge commit or empty child is not the permitted review-artifact child. Any implementation, unrelated documentation, cleanup, dependency change, other task commit, unrelated commit, or second post-review commit after R invalidates accepted lineage and requires a new report plus Architect review.

Canonical continuation does not derive a promotion branch from topology labels. Its expected_refs are zero or more explicit {ref, commit} identities, and canonical promotion uses PROMOTE_TARGET_REF plus promotion_target_ref; that target ref must be one of the expected refs so stale-target evidence is explicit. Existing expanded protocol-v3 expected_refs.dev/main plus PROMOTE_TO_MAIN remains compatibility input only and must not become the source of canonical branch-name assumptions.

Before promotion, refresh the candidate and the explicit promotion target ref, require the candidate ref equals the exact accepted candidate, require the target ref still equals its expected commit when that ref is part of the continuation snapshot, require authoritative verification to identify that exact SHA when required, require explicit promotion authority, and preflight promotion capability. If the candidate or required target ref changes after verification: REVERIFY / REVIEW_REQUIRED.

Do not auto-promote after push, review, CI, or verifier success. AUTO_UNTIL_STOP may only let orchestration dispatch an already-authorized promotion phase; it does not create promotion authority.

Release lifecycle

After exact promotion, derive PROMOTED_NOT_RELEASED until all separately authorized release actions are complete. Tag creation, repository metadata mutation, and GitHub Release publication are independent actions. Missing authority or capability stops the release phase without rolling back the validity of the completed promotion.

RELEASED requires completed authorized release actions plus final exact identity/state verification. Promotion and release are never implied by a previous phase.

GitHub Actions

For repositories that use DEV_MAIN, prefer one bounded push-to-integration validation workflow with relevant paths, one standard Linux job, read-only contents, immutable action pins, concurrency cancellation, short timeout, validator/tests, and no unnecessary dispatch, schedule, duplicate PR workflow, matrix, artifacts, cache, automatic rerun, larger runner, or paid external service. Other topologies use their target-authoritative integration/working branch rather than assuming dev.

Prefer one repository-owned deterministic verifier reused through whichever authorized surface is currently cheapest and sufficient. Local shell, Codex, Agent Runtime, and GitHub Actions should invoke the same verifier when feasible instead of carrying divergent verification logic. Independent deterministic local checks may be grouped into one bounded EXECUTION BUNDLE when the Executor/verification safety predicates hold; Git/ref publication remains serial. GitHub Actions is the bounded remote evidence boundary when exact task/target authority requires remote proof, not the default iterative debugger or mandatory development loop. ChatGPT+GitHub-only execution remains valid when no native-only capability is materially required.

Git success, CI success, Architect acceptance, authoritative project PASS, promotion authorization, actual promotion, release authorization, and actual release remain separate signals.

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.