oss-release-readiness
Skill0 starsChanged 10 days agoMITView on GitHub
Use before making a private repo public, open-sourcing a project, or pushing it to a public GitHub remote. Checks HEAD and all history for forbidden strings, author emails, secrets, sensitive files, notebook outputs, large files, missing LICENSE/README/SECURITY and third-party attribution.
- Reads credentials
In this repository20 other files it gives its agents.
- AGENTS.mdAGENTS.md
- examples/CLAUDE.mdCLAUDE.md
- api-and-interface-designSkill
- backlogSkill
- c4-architectureSkill
- change-recordSkill
- ci-cd-and-automationSkill
- competitive-briefSkill
- documentation-and-adrsSkill
- frontend-designSkill
- frontend-ui-engineeringSkill
- git-workflow-and-versioningSkill
- performance-optimizationSkill
- security-and-hardeningSkill
- shipping-and-launchSkill
- synthesize-researchSkill
- user-researchSkill
- ux-copySkill
- webapp-testingSkill
- web-visual-designSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
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 17
Secret scan: Run gitleaks git <repo> over the full history.
- Very specific: line 18
Manual review: Read third-party licences and each SOURCE.md (upstream commit SHA, modifications).
- Very specific: line 19
Back up: Before any rewrite make a full mirror: git clone --mirror <repo> <backup>.git.
- Distinctive: line 21
Re-check: Run steps 1 and 2 again on the rewritten repo.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
OSS Release Readiness
Publishing exposes the whole history, not just the latest files. A string removed in a later commit is still public. Check both.
Steps
Scripted check. Run it on the repo; each
--forbidvalue is a string that must not appear anywhere (personal email, home path, real name, internal host). Use the=form so a value starting with-is not read as an option. Each spelling needs its own--forbid=(a Windows path as backslash, forward slash and/c/...are three values). Under Git Bash setMSYS2_ARG_CONV_EXCL='*'first, or run from PowerShell, because Git Bash rewrites backslash paths. A bare user name can false-positive on third-party names; prefer the longer form.python -I scripts/release_check.py <repo> --forbid=<text> --forbid=<text>Exit 0 clean, 1 findings, 2 not a git repo, no commits, bad usage or a git failure. Findings:
forbid[n] content|history|commit-message|author-email|ref|tag|path-head|path-history(n is the index of the--forbidvalue; the value itself is never printed;content ... binarymeans a binary file matches;author-emailcovers author and committer name and email),shallow-clone(history is partial; fetch it in full and rerun),sensitive-path(.env.example,.env.sampleand.env.templateare not flagged),notebook-output,notebook-unreadable,large-file,missing,attribution.- Secret scan. Run
gitleaks git <repo>over the full history. Treat any hit as leaked: rotate the credential, do not just delete it. - Manual review. Read third-party licences and each
SOURCE.md(upstream commit SHA, modifications). Read README, SECURITY.md and docs for private names, internal URLs and anything the owner would not post publicly. - Back up. Before any rewrite make a full mirror:
git clone --mirror <repo> <backup>.git. Keep it private and outside the repo. - Rewrite history (owner only). By rule, the owner runs the rewrite in their own terminal, on a fresh clone. The tier hook does not list
git filter-repo, so nothing technical stops an agent: do not run it from a Claude session. Usegit filter-repoon a fresh clone:--mailmap <file>maps old author and committer identities to a public one,--replace-text <file>replaces forbidden strings in file contents with placeholders (old==>newper line), and--replace-message <file>does the same for commit messages (--replace-textdoes not touch them). Annotated tags and tagger emails need their own handling: delete and recreate the tags after the rewrite. List every spelling of a Windows path as a separate rule: backslash, forward slash, the/c/...form and the JSON-escaped form (\).--replace-textskips blobs with a NUL byte in the first 8 KB, so tracked binaries such as.pyckeep their strings: remove those paths with--invert-paths --path-glob '<glob>'and rerun the check. Remove sensitive paths the same way (--path <path> --invert-paths). Rewriting changes every commit hash, so tags and open work must be redone. - Re-check. Run steps 1 and 2 again on the rewritten repo. It must exit 0 before going on.
- Owner decision. The owner confirms the findings are resolved and makes the repo public. Do not change visibility or push to a public remote on their behalf.
Limits
- The check matches text case-insensitively, in text and binary files at HEAD (binary matches are reported as
binary), in diffs across history, commit messages, annotated tags and file names. Encoded or obfuscated secrets are not found. - Author and committer names and emails, ref names and annotated tags (message and tagger) are scanned.
- On exit 2 git's own error message is printed (truncated) and may contain a path; do not paste it into a public place without reading it.
- A bare or mirror clone returns 2; scan a normal clone.
- A directory nested inside another git repo scans that outer repo.
attributiononly checks thatskills/*/with aLICENSEhas aSOURCE.mdwith a 40-character upstream SHA; whether the licence permits redistribution is a human call.
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.

