pull-request
Skill1 starsChanged 2 months agoMITView on GitHub
Close a slice at an open draft pull request — read the diff, band its risk, and write the body a person merges from: Summary · Decided for you · Evidence · Risk band. Never merges, promotes to ready or deploys. Grading the diff is `code-review`; the release after a merge is `shipping-and-launch`.
- Commits and pushes
53 other projects use a version of this.
Adopted by
Most recently taken up by
- CVector-Energy/claude-code-solvetheir own version, first seen Nov 8, 2026
- neomjs/neo-agent-skillstheir own version, first seen Nov 8, 2026
- Kangentic/kangentictheir own version, first seen Nov 7, 2026
- speakeasy-api/gramtheir own version, first seen Nov 6, 2026
- StephanMeijer/skillstheir own version, first seen Nov 6, 2026
- drinkspiller/skillstheir own version, first seen Nov 6, 2026
Used with
Projects with it also use issue-campaign, contracts and 10 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.
samchonSkillissue-campaign
Defines the default solo repository-wide issue campaign for typia: exhaustive discovery, lead-vetted issue publication, one unified CI-validated implementation pull request per cycle, solo Self-Review, and repeated rediscovery until a full clean round. Use for broad audits, many issue candidates, or repeated issue-to-pull-request campaigns unless the user explicitly requests parallel or multi-agent execution; do not use for one already-defined issue or an ordinary pull request.
Also in 3 projects that use this
AnthropicSkillcontracts
Answer a question across a corpus of contract documents with verified citations. Use when the user asks what a contract says, which contracts have a clause, what changed between amendments, or any question that needs reading and citing across a set of contract files. The corpus must be on the local filesystem (see README).
Also in 3 projects that use this
mfang0126Skilldevelopment
Full-stack development workflow skill. Guides the complete cycle from understanding requirements to shipping code via PR with Preview validation.
Also in 3 projects that use this
dragonflydbSkillbenchmark
Benchmark Dragonfly (and compare against Valkey/Redis) on local or remote cloud instances, then produce a performance + memory report with charts and a raw-data appendix. Use this whenever the user runs /benchmark, or asks to benchmark / load-test / measure throughput, QPS, latency, or memory efficiency of Dragonfly or another Redis-compatible server, to compare Dragonfly vs Valkey or Redis, or to generate a benchmark report. Trigger it even when the user just describes the goal ("see how fast Dragonfly is on this box", "compare memory per key vs Valkey") without saying the word "benchmark".
Also in 4 projects that use this
VercelSkillgh-stack
Manages stacked PRs and splits multi-part work into reviewable branches with gh-stack. Use for stack creation, viewing, edits, push, submit, sync, rebase, merge, or checkout; when asked to split or isolate work for review; whenever a user mentions a stack, branch layers, dependent PRs, or gh stack; or when a stack is checked out.
Also in 3 projects that use this
carapace-shSkillgit
Use when working with Git — object model, references, index, branching, merging, rebasing, remotes, diffs, config, hooks, worktrees, submodules, and plumbing commands. Triggers on: "git", "git branch", "git commit", "git merge", "git rebase", "git checkout", "git stash", "git reset", "git submodule", "git worktree", "git blame", "git bisect", "git reflog", "git hooks", "git config".
Also in 3 projects that use this
Other versions53 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 53
- samchon/typia★ 6ksame name
- epicenter-md/epicenter★ 4.8ksame name
- EpicenterHQ/epicenter★ 4.8ksame name
- werf/werf★ 4.7ksame name
- basicmachines-co/basic-memory★ 4.5ksame name
- PaulWoitaschek/Voice★ 3.3ksame name
- event-catalog/eventcatalog★ 3ksame name
- heyitsnoah/claudesidian★ 2.6ksame name
- samchon/nestia★ 2.2ksame name
- theopenco/llmgateway★ 1.7ksame name
43 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository41 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- acceptance-criteriaSkill
- aim-higherSkill
- api-designSkill
- architecture-designSkill
- browser-testing-with-devtoolsSkill
- ci-cdSkill
- codebase-designSkill
- codebase-researchSkill
- code-reviewSkill
- code-simplificationSkill
- comprehension-quizSkill
- debugging-and-error-recoverySkill
- deprecation-and-migrationSkill
- documentation-and-adrsSkill
- doubt-driven-developmentSkill
- environment-manifestSkill
- gauntlet-loopSkill
- git-workflowSkill
- handoffSkill
- idea-refineSkill
- incremental-implementationSkill
- interview-meSkill
- literate-explainerSkill
- observability-and-instrumentationSkill
- orchestratorSkill
- performance-optimizationSkill
- plan-breakdownSkill
- preflight-readinessSkill
- project-setupSkill
- quality-verificationSkill
- security-and-hardeningSkill
- shipping-and-launchSkill
- source-driven-developmentSkill
- spec-grillingSkill
- spec-reviewSkill
- test-driven-developmentSkill
- to-prdSkill
- using-agent-skillsSkill
- worktreeSkill
The file
We split this skill into the rules it gives an agent
1 rule. 0 are common, 0 are distinctive, and 1 is very specific to this project.
0
Common
Most of the 48 edited versions we read keep these.
0
Distinctive
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- could be automatic: a setting could enforce it
- Very specific: line 104
Never gh pr merge, push to main, mark it ready, enable auto-merge or trigger a deploy — a person merges (safety-rails.md), and promoting …
“Common” is measured: kept by at least 6 in 10 of the 48 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.
Files
Pull request
Purpose
Stage: Ship. Principles 1, 7, 10.
Ends a slice at an open, risk-banded draft pull request: read the diff, band what it touches, write the body a person merges from. Their merge is the run's last independent gate, and the band is how they triage a queue without opening every diff cold.
When to use / when to skip
- A slice's code is committed on its branch and its gates have returned: the span ends here.
- The Ship step of a run, once per shipped slice; or by hand, on any branch to be opened as a draft.
- Skip where a stop-list condition fired
(
safety-rails.md): the slice ends blocked and nothing opens. - Near-miss — five-axis grading of the diff:
code-review. Scenarios against the running app:quality-verification. The release after a person's merge:shipping-and-launch. Branch and worktree lifecycle:git-workflow,worktree.
Inputs
Take each from the prompt where it is given, the canonical path otherwise. A person is present only when this runs by hand: at most three questions, and only where the gap changes the pull request's shape.
- the slice diff,
git diff <base>..HEADand its log — helps: what every step reads · without it: diff the working tree against the branch point, name the range,derived. docs/features/<slug>/qa.md— helps: the behavioural ledger, thenot-reachableids, the verdict · without it: build the test plan from the diff's tests and the commands you ran,derived.docs/features/<slug>/acceptance.md— helps: the signed scenario ids theqa.mdledger is checked against · without it: take the ledger's own id set,derived.docs/features/<slug>/prd.md## Solutionand the ADRs it cites — helps: the why the Summary carries · without it:intent.md, the plan, or the branch's purpose,derived.docs/features/<slug>/plan.mdandplan/<slice-id>.md— helps: the slice id, and the section each Decided-for-you row cites · without it:STATE.mdor the branch name,derived.docs/session-log.md— helps: what the run decided itself, and any oracle it changed · without it: reconstruct both from the diff and its commits,derived.
Process
- Read the diff first, and name the base ref you read it against. Every judgement below is a fact about the diff, never about the slice title or what the maker said it did.
- Check the stop list (
safety-rails.md): a secret in the diff or its history, or a Critical or High security finding, ends the slice there. Open no pull request and report which fired. Band the diff, and band it here only — every other file reproduces this band, because one computed two ways is a band nobody trusts. It is the higher of two readings.
Blast radius. Any row that fires makes it HIGH, and a row the diff cannot settle fires: nothing downstream re-reports an unraised band. Read what the diff touches before the ledger for how it passed — a clean record is what makes an auth or migration diff look safe.
The diff… changes authentication, authorization, session or permission logic changes payment, billing or pricing code reads, writes, exports or deletes personal data, or moves a schema holding it handles a credential — reads, stores, rotates or transmits a secret, token or key carries a destructive or irreversible migration — a dropped column, an unreplayable backfill, a bulk delete changes deploy, rollout or infrastructure configuration The quiet greens — where an unattended defect ships. Any one makes it MEDIUM, none makes it LOW: a
concernsverdict, anot-reachableid, coverage below the ledger's full set (exercised ÷ total —qa.mdreports no percentage), all three implement→verify→review rounds spent, an oracle the diff changed or removed, a value carried into the bodyderived, a diff adding a retry, a queue or an external call with no telemetry beside it — the path that goes quiet is the one nobody can debug from production.A HIGH band never holds, hides or halts a slice — it is how a person triages. Nor is it a High security finding: that one is the stop-list item, while correct auth work carries a clean verdict and a HIGH band.
- Pick the 3-5 files most likely to hide a subtle correctness defect — a new abstraction, dense logic, a side effect, the largest change — one line each on why. "All changed files" is the unfiltered diff nobody reads; a docs- or config-only slice names fewer rather than padding.
- Write the body as the brief a person merges from without re-reading the diff cold — the four
sections of Outputs & handoff, in order. Draw the Summary from
prd.md## Solutionand the ADRs by id, never a commit-log dump and never their reasoning restated. Transcribe theqa.mdledger rather than inventing a scenario↔test map, and give everynot-reachableid a human-ack checkbox — absorbing one defeats the run's only human-anchored oracle. Take Evidence's commands and their real output from the slice's last checkpoint andqa.md, re-running only what the diff moved. Narrate the diff layer to layer, never as a file list. Disclose any test, fixture or scenario the slice changed or removed — before, after, why; nothing is frozen, so the undisclosed change is the failure. - Give every in-run default its Decided-for-you row, matching one
docs/session-log.mdentry each. Where a plan section moved under a decision citing it, re-read and confirm or reverse that decision in the same pass — none lapses on its own. Push the branch, then open the draft — against the
<body>steps 5-6 wrote, never one still empty.git push -u origin <slice-branch> # never main gh pr create --draft --title "<type>(<scope>): <slice-id> <description under 70 chars>" --body-file <body>Never
gh pr merge, push tomain, mark it ready, enable auto-merge or trigger a deploy — a person merges (safety-rails.md), and promoting the draft spends the last reversible moment the work has. Return the URL; the push is yours, the worktree the caller's.
Rationalizations
| Rationalization | Reality |
| "The commit messages already explain it" | Commits carry what changed; the why lives in prd.md and the ADRs. |
| "Reviewers can open the files tab" | The unfiltered diff is what nobody reads; the curated 3-5 files is the deliverable. |
| "One scenario was unreachable, so leave it out" | An absorbed id is a gap the human-anchored oracle never sees. |
| "Everything is green, so open it ready to merge" | The draft is the last reversible moment; promoting it is the run grading itself. |
| "It touches auth, so hold it for a human first" | The band already flags it in that person's queue; holding takes it out. |
| "Every gate passed first time, so this is LOW" | Nothing in a slice's title or record says what its diff touches. |
Red flags
- A
blast radius:line missing or blank, which reads downstream as "none fired". - A code-reading checklist saying "all changed files", or a file listed with no reason beside it.
- A Summary assembled out of the branch's commit messages.
- A
not-reachableid that appears nowhere in the body. - A test or scenario changed inside the diff and mentioned in neither the body nor the log.
- A slice held back, or a pull request left unopened, because the band came out HIGH.
Verification
- One draft pull request is open on the slice branch,
mainuntouched, auto-merge off, URL returned. - The body carries
Summary · Decided for you · Evidence · Risk bandand nothing else, its word count measured against the 500-word cap. - The band is the higher of the two readings, its
blast radius:line naming every fired row with the file that fired it, or readingnone. - The checklist names 3-5 files with the reason each is there — or fewer, with the why stated.
- Every
not-reachableid carries a human-ack checkbox, every changed oracle its before, after and why, and every in-run default a row matching one log entry. - The secret scan came back clean, and Evidence quotes real build and test output.
- Every derived value reads
derived; nothing reconstructed is presented as signed. - Or: nothing opened, a stop-list condition having fired, and the report names which.
Outputs & handoff
The pull request body — cap 500 words, these four stable sections in this order and no others. Report the measured count against the cap; an over-cap body is trimmed before it opens, never handed on as if it fit.
## Summary why this slice exists — from prd.md ## Solution, ADRs by id, links to the artifacts
## Decided for you | # | Decision | Default taken | Why | Alternative | Plan section |
## Evidence the narrated diff · the transcribed qa.md ledger · [ ] a human-ack line per
not-reachable id · any oracle changed · the 3-5-file checklist · the commands run
with their real output · sibling slices still open or blocked
## Risk band LOW | MEDIUM | HIGH
blast radius: <none | auth: src/auth/session.ts · data: migrations/007_drop_email.sql>
quiet greens: verdict · exercised/total · not-reachable ids · rounds · oracle changedRename one of the four and its consumer changes in the same commit: a body whose shape moved silently is read by a person expecting the old one.
Appended to docs/session-log.md: one entry, cap 60 words, where this skill settled something itself
— the base ref, a derived slice id. Shape and append rules:
state-schema.md.
Returned in conversation: the pull request URL and its band — or the stop-list condition that fired, and no pull request.
Nothing else. No STATE.md row, the caller owning the slice's transition
(state-schema.md); no qa.md, no lesson, no release notes, no merge.
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
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.

