supply-chain
Skill0 starsChanged 47 days agoMITView on GitHub
Adding or upgrading a dependency, or reviewing a lockfile change. Vetting, pinning, provenance, known vulns, typosquats, CI token scope.
- Installs packages
9 other projects use a version of this.
Adopted by
Most recently taken up by
- Atmoskun/claude-skillstheir own version, first seen Nov 16, 2026
- mrhlaingbwardev/agentic-hacking-skillstheir own version, first seen Nov 8, 2026
- nimadorostkar/Claude-Skills-collectiontheir own version, first seen Oct 13, 2026
- nitinjain999/platform-skillstheir own version, first seen Oct 13, 2026
- padamson/playwright-rusttheir own version, first seen Oct 8, 2026
- daloopa/investingtheir own version, first seen Oct 8, 2026
Other versions9 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 9
- daloopa/investing★ 488same name
- mrhlaingbwardev/agentic-hacking-skills★ 347same name
- padamson/playwright-rust★ 169same name
- nitinjain999/platform-skills★ 71same name
- nimadorostkar/Claude-Skills-collection★ 37same name
- thejefflarson/soundcheck★ 20same name
- poorvith-mp/skills-business★ 1same name
- Atmoskun/claude-skills★ 0same name
- conjure-3301/skills★ 0same name
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 repository29 other files it gives its agents.
- hosts/AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- adrSkill
- api-designSkill
- auditSkill
- code-reviewSkill
- concurrency-performanceSkill
- coverageSkill
- debuggingSkill
- debugSkill
- fast-laneSkill
- fixSkill
- implementSkill
- intakeSkill
- leanSkill
- migration-safetySkill
- nonnaSkill
- observabilitySkill
- planSkill
- refactoringSkill
- releaseSkill
- reviewSkill
- rollbackSkill
- security-reviewSkill
- shipSkill
- syncSkill
- tddSkill
- tdd-workflowSkill
- testSkill
The file
We split this skill into the rules it gives an agent
5 rules. 0 are common, 1 is distinctive, and 4 are very specific to this project.
0
Common
Most of the 8 edited versions we read keep these.
1
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 14
Run the bundled scripts/dep-audit.sh on any dependency or lockfile change — it auto-detects the ecosystem and runs the right vuln scanner, …
- Very specific: line 27
What does it drag in?: Look at the transitive tree and the install scripts.
- Distinctive: line 29
Read it. Prefer dependencies with few or no transitive deps and no install hooks.
- Very specific: line 59
Run the ecosystem scanner in CI and fail the build on findings: npm audit, pip-audit, govulncheck, cargo audit (the bundled dep-audit.sh …
- Very specific: line 69
Verify the exact name: before installing: requets, python-dateutil vs python3-dateutil, crossenv vs cross-env.
“Common” is measured: kept by at least 6 in 10 of the 8 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
Supply Chain Security
Your dependencies run with your code's full privileges — in your CI, your build, and production. A
compromised package is a compromise of you. Most breaches in this space aren't exotic: a typosquat
someone installed, a postinstall script no one read, a known CVE that sat un-upgraded, or a CI token
with far too much scope. The discipline is distrust by default and verify before adopting
(.claude/rules/boundaries.md, .claude/rules/safety.md).
Run the bundled scripts/dep-audit.sh on any dependency or lockfile change — it auto-detects the
ecosystem and runs the right vuln scanner, failing on findings.
Before you add it — vetting a new dependency
Adding a dependency is a permanent liability and a new entry in your attack surface. The bar is high.
- Do you need it at all? The Nonna rule: reuse a maintained library when it covers ≥80% of the
need — but a one-liner you can write and own beats pulling a transitive tree for a trivial
helper (
left-pad). For parsing, dates, crypto, auth — always the vetted library, never hand-rolled (.claude/rules/dev-process.md,.claude/rules/engineering.md). - Is it healthy? Recent commits, real maintainers (bus factor > 1), responsive security history, adoption you can verify. An abandoned package with a known CVE is a time bomb.
- What does it drag in? Look at the transitive tree and the install scripts. A package
with a
postinstall/preinstallruns arbitrary code on everyinstall— in CI, with your tokens. Read it. Prefer dependencies with few or no transitive deps and no install hooks. - What can it reach? A library that wants network/filesystem/child-process it has no business needing is a red flag. Least privilege applies to code you import, not just code you write.
Capture the why of a non-obvious adoption in an ADR (/adr) — the next person will ask.
Pinning & lockfiles — reproducible by construction
A build that resolves different bytes on different days is unauditable and un-reproducible
(.claude/rules/boundaries.md).
- Commit the lockfile (
package-lock.json,poetry.lock,Cargo.lock,go.sum,pnpm-lock,requirements.txtwith hashes). It is the source of truth for exact resolved versions. - Install from the lock, frozen, in CI:
npm ci(notnpm install),poetry install --sync,pip install --require-hashes,go mod verify,cargo build --locked. A CI that re-resolves can silently pull a new (or malicious) version — pin and freeze. - Pin exact versions for apps; a floating
^/~/*range means your next build is a gamble on whatever was published since. (Libraries publish ranges; applications pin.)
Provenance & integrity — verify the bytes are the ones you vetted
- Hash-locked installs: the lockfile records an integrity hash per package; the frozen install verifies it. Don't disable integrity checking to make an install pass — that's removing the gate.
- Prefer signed/attested artifacts where the ecosystem supports it (Sigstore/cosign, npm
provenance,
go.sum's checksum DB). Provenance ties the artifact to its source build. - Pin CI actions/images by digest, not a moving tag.
actions/checkout@<sha>not@v4;image@sha256:...not:latest. A tag can be re-pointed at malicious code under you.
Known-vuln scanning — automated and blocking
- Run the ecosystem scanner in CI and fail the build on findings:
npm audit,pip-audit,govulncheck,cargo audit(the bundleddep-audit.shpicks the right one). A vuln report that doesn't block merge is a report nobody reads. - Triage, don't blanket-ignore. Suppress a specific advisory only with a written reason and an
expiry, never a global
--audit-level=none. Fix-by-upgrade beats suppress. - Scan on a schedule, not just on change — CVEs are disclosed against versions you already ship. A dependency safe at merge is vulnerable the day an advisory drops.
Typosquatting & install-time attacks — the cheap, common exploit
- Verify the exact name before installing:
requets,python-dateutilvspython3-dateutil,crossenvvscross-env. A transposed letter installs an attacker's package that runs on install. - Watch for dependency confusion: a public package shadowing your internal name. Scope/namespace internal packages and configure the registry to prefer your private source.
- Audit the lockfile diff, not just
package.json. The danger hides in transitive resolutions and integrity-hash changes a top-level bump pulled in. A one-line manifest change can rewrite dozens of resolved packages — review them.
Least-privilege CI tokens — assume the build can be compromised
CI is where supply-chain attacks cash out: it has your publish tokens, cloud creds, and secrets.
- Scope tokens to the minimum and prefer short-lived OIDC federation over long-lived static secrets. A read-only job gets a read-only token.
- Don't expose secrets to PRs from forks, and don't let untrusted PR code run in a context that
can read them (the classic
pull_request_targetfoot-gun). - Pin and minimize the actions/plugins your pipeline runs — each is code with access to that token. Same vetting bar as a runtime dependency.
Reproducible builds
Same source + same locked deps → bit-identical artifact. Pin the toolchain/base image by digest,
freeze deps from the lock, avoid embedding timestamps/absolute paths/build-host state. Reproducibility
is what lets you prove the artifact you're shipping is built from the source you reviewed — without it,
"it's the same code" is a hope, not a fact (.claude/rules/boundaries.md).
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.

