accessibility-complian…
Web accessibility patterns for news and academic sites. Use for WCAG audits, alt text,…
Local Gitleaks scans for staged changes, push ranges, and full history in private repos, with redacted reports. Use to block committed secrets without GitHub Secret Protection.
$ npx -y skills add jamditis/claude-skills-journalism --skill private-secret-scanning --agent claude-codeHow it fires
How this skill gets triggered: by you, by Claude, or both.
/private-secret-scanningContext preview
The summary Claude sees to decide when to auto-load this skill.
Local Gitleaks scans for staged changes, push ranges, and full history in private repos, with redacted reports. Use to block committed secrets without GitHub Secret Protection.
name: private-secret-scanning description: Local Gitleaks scans for staged changes, push ranges, and full history in private repos, with redacted reports. Use to block committed secrets without GitHub Secret Protection.
GitHub push protection and secret scanning are free on public repositories. Private repositories need a paid GitHub Secret Protection license. This skill gives private and shared repositories a deterministic local replacement: a pinned [Gitleaks](https://github.com/gitleaks/gitleaks) binary, three scan scopes, and git hooks that stop a leak before it leaves the machine.
Everything runs through one script that ships with this skill: `scripts/secret-scan.sh`. It needs only Bash (3.2 or later, so the macOS default works), git, curl, and python3.
SCAN=path/to/private-secret-scanning/scripts/secret-scan.sh "$SCAN" install # pinned Gitleaks 8.30.1, SHA-256 checked before install "$SCAN" self-test # proves leaks fail and clean repos pass on this machine cd your-repo "$SCAN" history --report /tmp/secret-report.json # one-time audit of every commit "$SCAN" install-hooks # pre-commit and pre-push protection from now on
| Command | Scans | Use it | |---|---|---| | `staged` | the staged diff only | pre-commit hook; fastest | | `push [RANGE]` | commits in `RANGE`, or the commits a `git push` is about to send (read from pre-push stdin). A new branch or tag is scanned through its full history | pre-push hook; catches commits made with `--no-verify` | | `history` | every commit reachable from any ref, including files deleted later | first adoption, and before a private repo goes public |
A secret that was committed and then deleted is still in history. Only `history` finds it, and only rotating the credential fixes it. Rewriting history does not undo a clone that already happened.
| Exit | Meaning | |---|---| | 0 | no leaks | | 1 | leaks found; each is listed on stderr | | 2 | scanner or configuration error |
The script fails closed. It exits 2, never 0, when:
A hook that exits non-zero blocks the commit or push, so a broken scanner blocks work. It does not let a leak through.
Terminal output and `--report` files hold five fields per finding: `RuleID`, `File`, `StartLine`, `Commit`, and `Fingerprint`. They never hold the secret, the match, or the source line. Gitleaks runs with `--redact`, its raw report goes to a private temp file that is deleted on exit, and the script copies only those five fields out. The report is written with mode `0600`.
The fingerprint (`commit:file:rule:line`) is enough to find and fix the leak. It is also the line you add to `.gitleaksignore` to accept a finding.
Adopting scanning on an old repo usually turns up findings you have already rotated. Record them instead of turning the scanner off:
1. Rotate every live credential first. Accepting a finding does not make it safe. 2. Run `history --report /tmp/secret-report.json`, then write its fingerprints to `.gitleaksignore` at the repo root:
python3 -c 'import json,sys; print("\n".join(f["Fingerprint"] for f in json.load(open(sys.argv[1]))))' \
/tmp/secret-report.json >> .gitleaksignore3. Commit `.gitleaksignore`. Later scans report only new findings. Each line names one commit, file, rule, and line, so a new leak of the same secret in another commit still fails.
A `.gitleaks.toml` replaces the built-in rules unless it extends them. An allowlist-only file therefore turns every detector off, and the script refuses one. Start every config from this:
[extend] useDefault = true [allowlist] description = "synthetic keys in test fixtures" paths = ['''^tests/fixtures/''']
Keep allowlists narrow. Allow a specific path and rule, such as a test fixture, instead of a whole directory or a broad regex. Never allowlist a rule wholesale because it is noisy. A config that silences a rule also silences the next real leak of that type. Review a `.gitleaksignore` or allowlist change the way you would review an access-control change.
The script picks up `.gitleaks.toml` and `.gitleaksignore` from the repo root automatically. Hooks use them only once they are committed (or staged, for `staged`). A local edit cannot quietly change what gets through. Set `SECRET_SCAN_CONFIG` to use a shared config from elsewhere; a relative path is resolved from where you run the script.
`install-hooks` writes `pre-commit` (runs `staged`) and `pre-push` (runs `push`) into the repository's hooks directory. It refuses to overwrite a hook it did not write; add a call to the script inside that hook instead. `git commit --no-verify` skips pre-commit, which is why pre-push scans the outgoing range again. A new branch is scanned through its full history, even if a remote-tracking ref says a remote already has those commits. Tracking refs can be stale: `git remote set-url` points a remote at a new URL, and the old refs stay. This matters when a branch moves from a private remote to a public one.
A collection of Agent Skills for journalists, researchers, academics, media professionals, and communications practitioners. The same repository serves Claude Code and Codex while keeping Claude-only commands, agents, and hooks clearly labeled.
Repo: jamditis/claude-skills-journalism
Web accessibility patterns for news and academic sites. Use for WCAG audits, alt text,…
Scans the session for lessons and workflows, then proposes scoped CLAUDE.md edits. Use for…
Manages attention and evidence in long agent sessions. Use for lost instructions, dropped…
Directs the current request through configured lower-tier agents. Use only for explicit…
Electron desktop apps with React, TypeScript, and Vite. Use for IPC, window/tray, PTY…
Remote JavaScript console and debugging on mobile. Use for phone/tablet console errors,…