implement
Skill4 starsChanged 57 days agoApache-2.0View on GitHub
Work a GitHub issue end to end on its own branch and open a pull request against edge. Use for /impl, /implement, or when asked to start work on an issue number.
- Commits and pushes
192 other projects use a version of this.
Adopted by
Most recently taken up by
- philoserf/claude-code-configtheir own version, first seen Nov 15, 2026
- muxy-app/muxytheir own version, first seen Nov 15, 2026
- geoffbremner/skillstheir own version, first seen Nov 15, 2026
- SireJeff/k0ntexttheir own version, first seen Nov 15, 2026
- quickchatai/quickreviewtheir own version, first seen Nov 15, 2026
- HiteSit/real-scientist-skillstheir own version, first seen Nov 15, 2026
Used with
Projects with it also use setup-ts-deep-modules, ask-matt 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.
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
- +3 more from google
Other versions192 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 192
- mattpocock/skills★ 325ksame name
- QuivrHQ/quivr★ 40ksame name
- The-Vibe-Company/quivr★ 40ksame name
- foambubble/foam★ 17ksame name
- formbricks/formbricks★ 13ksame name
- ArcReel/ArcReel★ 6.6ksame name
- vinvcn/mattpocock-skills-zh-CN★ 5.8ksame name
- fallow-rs/fallow★ 5.3ksame name
- withcoral/coral★ 4.9ksame name
- bestofjs/bestofjs★ 3.1ksame name
182 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 repository14 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- commits-and-prsSkill
- decomposeSkill
- github-boardSkill
- releaseSkill
- reviewSkill
- reviseSkill
- rubber-duckSkill
- specifySkill
- thunderstormSkill
- triageSkill
- worktreeSkill
- writing-docsSkill
The file
We split this skill into the rules it gives an agent
9 rules. 0 are common, 6 are distinctive, and 3 are very specific to this project.
0
Common
Most of the 174 edited versions we read keep these.
6
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
- Distinctive: line 23
Stop and ask when any of these is true:
- Distinctive: line 31
Read CLAUDE.md and the spec the issue points at.
- Very specific: line 56
Every branch an agent pushes carries the agent/ prefix, which is what tells a human reading the branch list what an agent owns.
- Very specific: line 78
Use RUSTC_WRAPPER=sccache when it is installed.
- Distinctive: line 85
Write the test before or with the change.
- Distinctive: line 93
Run the narrowest relevant test, then the full gates:
- Very specific: line 101
Run pnpm --dir docs build when anything under docs/ changed.
- Distinctive: line 115
Do not weaken, skip, or delete a test to make a gate pass.
- Distinctive: line 156
Leave a worktree behind after the pull request merges.
“Common” is measured: kept by at least 6 in 10 of the 174 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
Take one issue, do the work on its own branch, open a pull request. One issue per run.
Read before claiming
The github-board skill's Transport section decides whether this run uses gh
or the GitHub MCP tools. Use the same transport for everything below.
gh issue view <n> --json title,body,labels,assignees,url
gh issue view <n> --json subIssuesThrough MCP: issue_read method get, then method get_sub_issues.
Stop and ask when any of these is true:
- The issue has no acceptance criteria you could verify.
- The issue has sub-issues. It is what
/stormtakes rather than a unit of work; run it, or work its sub-issues one at a time. - The issue is already
claimedby someone else. - The work needs a decision the issue does not record.
Read CLAUDE.md and the spec the issue points at. Read the code the change
touches before planning the change.
Claim it
Assign yourself and move the status label to claimed. The claim is what stops
a second run from taking the same issue, so make it before touching a tree.
Take a working tree
A dispatched implementer already has one: the run created it and handed over the directory. Use it and skip this section.
A session invoked directly takes its own tree. The worktree skill's Who gets
one section says which kind, and is the only thing that creates a worktree.
A cloud session holds a container checkout nobody else owns, so it works in that checkout and creates nothing beside it:
git fetch origin
git branch -m agent/<n>Rename the harness-supplied branch rather than keeping it or branching afresh.
Every branch an agent pushes carries the agent/ prefix, which is what tells a
human reading the branch list what an agent owns. The rename also keeps whatever
that branch already carries: git switch -c agent/<n> origin/edge matches it
only while the branch sits at origin/edge, and otherwise leaves its commits on
a name this session no longer uses. Harness branches get deleted, so that
residue becomes unreachable rather than untidy.
Where there is no branch to rename, that same git switch -c creates one.
A local session takes a worktree, because the checkout there is the user's. It
goes outside the repository root so Cargo does not find the parent
.cargo/config.toml:
git fetch origin
git worktree add ../thndrs-worktrees/<n> -b agent/<n> origin/edgeWork only inside that worktree. The user's primary checkout stays untouched.
Each worktree keeps its own target/. Do not set a shared CARGO_TARGET_DIR:
Cargo locks the build directory and a shared one serializes parallel builds.
Use RUSTC_WRAPPER=sccache when it is installed.
Do the work
Make the smallest change that satisfies the acceptance criteria. Follow the
module order, error policy, and documentation rules in CLAUDE.md.
Write the test before or with the change. A change with no test needs a reason in the pull request body.
Work that turns out to be larger than the issue describes is a signal, not a license. File the extra work as a new issue and finish what was claimed.
Verify
Run the narrowest relevant test, then the full gates:
cargo fmt
cargo clippy --workspace --all-targets --all-features --locked -- -D warnings
cargo test --workspace --all-features --lockedRun pnpm --dir docs build when anything under docs/ changed.
A change under cli/renderer/ or cli/app/ gets a captured frame:
cargo build -p thndrs && .claude/scripts/tui-capture.sh <scenario>Show the rendered image in your report. A capture nobody looks at has checked
nothing, so put the frame in front of whoever asked before calling the change
verified, and check any claim you make about it against the .ansi beside it. Two of the first three observations made from one
of these images were wrong about what the application drew.
docs/src/content/docs/docs/development/tui-qa.md carries the rest.
Do not weaken, skip, or delete a test to make a gate pass. If a test is wrong, say so in the pull request body and explain why.
Open the pull request
Base the pull request on edge.
The title and body become the squash commit, verbatim. Write them to that:
- Title under 53 characters, as
<type>: <what changed>. GitHub appends(#NN), and 59 is the limit. - Body under 20 lines, wrapped at 72 columns, with no headings.
## Whatreachesgit logas the literal characters## What. Closes #<n>, aVerified with <command>line naming a command you ran, and aNot coveredline that is not empty.
commits-and-prs carries the rest, and
.claude/scripts/check-commit-message.py --pr reports both before the merge.
Push with .claude/scripts/push-verified.sh, which compares the remote ref to
local HEAD afterwards. git push exits zero for a push that carried nothing,
so its exit code is not evidence that the branch moved.
gh pr create --base edge --head agent/<n> --title <title> --body-file <file>
# MCP: create_pull_request with base "edge", head "agent/<n>", and the body inline.Move the issue status to review.
Report
State what changed, what you verified and how, what you did not verify, and anything you filed as a separate issue. Do not claim a check passed without having run it.
Do not
- Merge, approve, or push to
edgeormain. - Touch the user's primary checkout on a development machine.
- Expand scope past the claimed issue.
- Leave a worktree behind after the pull request merges.
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.

