implement
Skill0 starsChanged 3 months agoMITView on GitHub
Implement a piece of work based on a PRD or set of issues.
191 other projects use a version of this, 64 of them edited.
Adopted by
Most recently taken up by
- muxy-app/muxytheir own version, first seen Nov 15, 2026
- geoffbremner/skillsedited, first seen Nov 15, 2026
- SireJeff/k0ntexttheir own version, first seen Nov 15, 2026
- quickchatai/quickreviewedited, first seen Nov 15, 2026
- HiteSit/real-scientist-skillsedited, first seen Nov 15, 2026
- felipeam86/pipe-skillstheir own version, first seen Nov 15, 2026
Used with
Projects with it also use setup-ts-deep-modules, ask-matt and 9 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.
liufeng1976Skillsetup-ts-deep-modules
Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.
Also in 43 projects that use this
GoogleSkillask-matt
Ask which skill or flow fits your situation. A router over the skills in this repo.
Also in 91 projects that use this
RoentekSkillloop-me
Grill me about specs for the workflows I want to build, within this workspace.
Also in 50 projects that use this
AWSSkillto-questionnaire
Turn a decision you can't fully answer into a questionnaire for someone else to fill in.
Also in 88 projects that use this
ryotaroh180105Skillclaude-handoff
Hand the current conversation off to a fresh background agent that picks up the work immediately.
Also in 42 projects that use this
GoogleSkillto-spec
Turn the current conversation into a spec and publish it to the project issue tracker — no interview, just synthesis of what you've already discussed.
Also in 168 projects that use this
- +2 more from google
Other versions191 projects use a version of this.
Word for word · 0
None found.
Edited · 64
- mattpocock/skills★ 325k19% same
- QuivrHQ/quivr★ 40k29% same
- The-Vibe-Company/quivr★ 40k29% same
- formbricks/formbricks★ 13k19% same
- ArcReel/ArcReel★ 6.6k19% same
- withcoral/coral★ 4.9k19% same
- bestofjs/bestofjs★ 3.1k29% same
- moeru-ai/xsai★ 66729% same
- aws/graph-explorer★ 49229% same
- biersoeckli/QuickStack★ 36519% same
54 more not listed
Wrote their own version · 127
- foambubble/foam★ 17ksame name
- vinvcn/mattpocock-skills-zh-CN★ 5.8ksame name
- fallow-rs/fallow★ 5.3ksame name
- huggingface/OpenEnv★ 2.9ksame name
- meta-pytorch/OpenEnv★ 2.8ksame name
- oxsecurity/megalinter★ 2.6ksame name
- muxy-app/muxy★ 2.4ksame name
- pymc-labs/pymc-marketing★ 1.3ksame name
- adobe/S3Mock★ 1.2ksame name
- nodetool-ai/nodetool★ 637same name
117 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 repository28 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- codebase-designSkill
- code-reviewSkill
- commission-method-gap-researchSkill
- diagnosing-bugsSkill
- domain-modelingSkill
- grillingSkill
- grill-with-docsSkill
- handoffSkill
- improve-codebase-architectureSkill
- method-gap-research-statusSkill
- prototypeSkill
- refactor-skillSkill
- research-briefSkill
- researchSkill
- setup-matt-pocock-skillsSkill
- skill-evidence-captureSkill
- skill-evolutionSkill
- skill-evolution-statusSkill
- tddSkill
- teachSkill
- to-issuesSkill
- to-prdSkill
- triageSkill
- wait-whatSkill
- wayfinderSkill
- writing-great-skillsSkill
The file
We split this skill into the rules it gives an agent
6 rules. 0 are common, 3 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 172 edited versions we read keep these.
3
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
- Very specific: line 11
Read the repository's agent instructions and the authoritative PRD or live issue bodies and comments.
- Very specific: line 18
Use /tdd where possible, at pre-agreed seams.
- Distinctive: line 22
Run the smallest applicable checks regularly, and run the full applicable verification set once at the end.
- Very specific: line 26
Put only the scoped implementation into HEAD (a provisional commit is acceptable) before invoking /code-review, because that skill reviews …
- Distinctive: line 34
Only when the user expressly authorizes implementation-closeout tracker mutations, follow the repository's tracker contract.
- Distinctive: line 40
Keep closeout evidence on one of two paths:
“Common” is measured: kept by at least 6 in 10 of the 172 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
Implement the work described by the user in the PRD or issues.
Before editing
- Read the repository's agent instructions and the authoritative PRD or live issue bodies and comments. For tracker reads and mutations, follow
docs/agents/issue-tracker.md§Conventions rather than copied summaries or duplicated mechanics. - Record the review fixed point the user supplies. If none was supplied, ask the user to confirm the current
HEADbefore editing. - Run
git status --shortwithout a path filter. Classify all pre-existing changes, preserve unrelated dirt, and keep the implementation scope explicit. Uncommitted state is unrecoverable and no later reconciliation restores it, so for the rest of the run every prompt you dispatch to a delegated agent must forbid the git commands that mutate the working tree or index —add,commit,checkout,restore,switch,stash,reset,clean,apply. Telling a delegate not to edit files does not cover them. - Turn the acceptance criteria into a run sheet that maps each requirement to its implementation seam and verification. When
/tddapplies, point each applicable requirement to its TDD evidence row instead of copying the TDD-specific fields into this run sheet.
Build and verify
Use /tdd where possible, at pre-agreed seams.
Discover the repository's native verification surface before selecting checks. Run typechecking, focused tests, and a full test suite only when they are configured and applicable. In prose or no-build repositories, use the checks named by the repository instructions, such as single-home, cross-reference, and release-metadata checks.
Run the smallest applicable checks regularly, and run the full applicable verification set once at the end. If no automated harness exists, record the exact structural or manual checks and their results.
Review and commit
Put only the scoped implementation into HEAD (a provisional commit is acceptable) before invoking /code-review, because that skill reviews committed HEAD against the recorded fixed point. Its reviewers are delegated agents, so item 3's delegation prohibition binds every packet dispatched to them.
Immediately before every provisional or final commit, or any amend, rerun unscoped git status --short, reconcile every transition against the initial classification, and inspect git diff --cached --name-only plus git diff --cached --check; proceed only when the index is exactly the scoped implementation and unrelated dirt remains unstaged. Immediately before /code-review, repeat the unscoped reconciliation, require an empty index, and inspect git diff --name-only <fixed-point>...HEAD plus git diff --check <fixed-point>...HEAD so the committed review range is exactly scoped.
Resolve every actionable Standards and Spec finding. If a finding would change a ratified interface, ownership boundary, scope, reversibility, pattern, or dependency, pause and re-ratify it with the user. When a finding changes observable behavior, re-enter /tdd and satisfy that skill's review-reentry evidence rule before applying the production repair; behavior-neutral refactors stay in this repair loop without inventing a red cycle. After each repair, commit or amend the scoped repair into HEAD, rerun applicable verification and /code-review, and continue until both axes are green on the exact current SHA. That green-reviewed SHA is the final implementation commit. Any later commit, amend, or other HEAD movement requires applicable verification and both review axes to run again against the replacement SHA before the workflow is treated as green.
Optional tracker closeout
Only when the user expressly authorizes implementation-closeout tracker mutations, follow the repository's tracker contract. Verify every requested dependent or child issue is closed before closing its parent PRD, then exact-read every requested state.
For work originating from a live issue, record an explicit closeout disposition before final handoff. If no disposition has been ratified once the green-reviewed SHA exists, perform the read-only repository, published-ref, and publication-delta checks below, disclose the exact publication set and intended tracker mutations, and ask the user to authorize any required publication plus closeout or to leave the tracker open. Treat broad closeout permission received before the publication set is known as closeout intent, not authorization for an undisclosed set; disclose the complete set and require confirmation before publishing it. If closeout remains unauthorized, leave the issue open and make its exact tracker state and the missing authorization prominent in the final response.
Before the first closeout mutation, identify the target repository and intended published ref, refresh or query its remote state, and prove that the green-reviewed implementation commit is reachable from that published ref; local branch or upstream status alone is not publication proof. Before asking for publication authorization, enumerate the exact commit delta from the queried published-ref SHA to the green-reviewed SHA (for example, git log --oneline <published-ref-sha>..<reviewed-head-sha>). Disclose every commit publication would make reachable, explicitly identify any commit outside the current implementation scope, and require authorization for the complete publication set. If authorization does not cover that full set, stop for a different ref or branch, or revised direction. If reachability is absent or cannot be established, stop and ask the user either to publish or authorize publication, or to explicitly authorize local-only closeout. Close permission does not grant push or publication authority. Disclose any local-only exception in both the tracker comment and final response.
Keep closeout evidence on one of two paths:
- External-only (default): preserve tracker comments and exact-read results outside the committed implementation, and make no tracked run-sheet or ledger change after the green-reviewed, published implementation commit.
- Committed evidence: if closeout must update a tracked run sheet or ledger, record that as a separate closeout-only delta after the tracker mutation; do not amend the already-published implementation commit. Rerun applicable verification and
/code-review, finalize the delta, apply the same publication-reachability gate to the resulting final workflow commit, and exact-read every requested tracker state before reporting the final SHA.
This closeout is separate from issue publication: /to-issues never closes a parent, and approval to publish slices does not authorize implementation closeout. Never infer close permission from implementation completion alone.
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.

