agentleFS
RealDougEubanksRealDougEubanks/ClaudeMarketplace/skills/pre-commit/SKILL.md

pre-commit

Skill1 starsChanged 2 months agoMITView on GitHub

Fast pre-commit quality gate: scans staged files for secrets, dead code, naming issues, merge conflict markers, and direct-to-main commits. Installs as a git hook via /pre-commit install.

11 other projects use a version of this.

Adopted by

Most recently taken up by

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. 0 are common, 6 are distinctive, and 10 are very specific to this project.

  • 0

    Common

    Most of the 11 edited versions we read keep these.

  • 6

    Distinctive

    This author's own choices. Worth a look.

  • 10

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    When invoked without arguments, run immediately against staged files (or all modified files if nothing is staged):

    line 18
  2. Very specific:

    Use Bash to get the list of files to check.

    line 20
  3. Distinctive:

    If both lists are empty, report "Nothing staged or modified — nothing to check." and stop.

    line 28
  4. Distinctive:

    If on main, BLOCKER: "You are committing directly to main.

    line 54
  5. Distinctive:

    If any BLOCKER exists: exit with a non-zero message and do NOT proceed with the commit.

    line 73
  6. Distinctive:

    If only WARNINGs: inform the user and ask "Proceed with commit anyway?

    line 74
  7. Distinctive:

    If all PASSED: confirm "All checks passed.

    line 75
  8. Very specific:

    When invoked as /pre-commit install:

    line 81
  9. Very specific:

    Use Bash to confirm this is a git work tree and resolve the hooks directory.

    line 88
  10. Very specific:

    Any other hook: if $HOOKS_DIR/pre-commit.backup.pre-commit already exists, stop and tell the user.

    line 99
  11. Very specific:

    Use mkdir -p "$HOOKS_DIR", then Write to create $HOOKS_DIR/pre-commit with the canonical script below verbatim — do not regenerate, …

    line 104
  12. Very specific:

    Use Bash to make the hook executable: chmod +x "$HOOKS_DIR/pre-commit"

    line 193
  13. Very specific:

    Use Bash to resolve the hooks directory: HOOKS_DIR=$(git rev-parse --git-path hooks).

    line 206
  14. Very specific:

    If $HOOKS_DIR/pre-commit does not exist, report that no hook is installed and stop.

    line 207
  15. Very specific:

    Check for the marker: grep -qx '# claude-skill-hook: pre-commit' "$HOOKS_DIR/pre-commit".

    line 208
  16. Very specific:

    If $HOOKS_DIR/pre-commit.backup.pre-commit exists, restore it as the active hook so the user's original hook keeps running:

    line 209

“Common” is measured: kept by at least 6 in 10 of the 11 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.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

pre-commit

Invoked via /pre-commit. This skill acts as a pre-commit quality gate — it runs a fast subset of checks before committing and blocks the commit if critical issues are found. It can also install itself as a git hook.


Mode A — Run checks now (default)

When invoked without arguments, run immediately against staged files (or all modified files if nothing is staged):

  1. Use Bash to get the list of files to check. git diff exits 0 even when it prints nothing, so an || fallback would never fire; test for empty output explicitly:
    FILES=$(git diff --cached --name-only --diff-filter=ACMR)
    if [ -z "$FILES" ]; then
      FILES=$(git diff --name-only --diff-filter=ACMR)
    fi
    printf '%s\n' "$FILES"
    If both lists are empty, report "Nothing staged or modified — nothing to check." and stop.
  2. Secret scan — Use Grep across all staged files for hardcoded secret patterns:
    • Assignments (= or : — covers code, YAML, and JSON): (?i)(password|passwd|pwd|api_key|apikey|api_token|access_token|secret_key|client_secret|secret)\s*[=:]\s*['"][^'"]{8,}['"]
    • Shell exports: export\s+[A-Z_]*(KEY|TOKEN|SECRET|PASSWORD)[A-Z_]*=\S{8,}
    • AWS: (?i)(aws_access_key_id|aws_secret_access_key)\s*[=:]\s*['"][^'"]{16,}['"] and literal AKIA[0-9A-Z]{16}
    • Modern token prefixes: sk-ant-, sk-proj-, github_pat_, ghp_, glpat- (GitLab), npm_ (npm), AIza (Google API key), xox[baprs]-, sk_live_
    • Private key headers: -----BEGIN (RSA|DSA|EC|OPENSSH) PRIVATE KEY-----
    • High-entropy heuristic: any assignment of a 32+ char mixed-case alphanumeric literal to a variable whose name contains key, token, secret, or auth
    • Exclude: *.example, *.sample, node_modules/, vendor/
    • BLOCKER if found, except in test files (*.test.*, *.spec.*), where a match is a WARNING: fixtures usually hold fake values, but real credentials do get pasted into tests, so they are still scanned and reported. Report file and line number. When reporting, redact the matched value (show the first 4 characters only).
  3. Dead code scan — Use Grep across staged source files for:
    • ^\s*//.* blocks of 3+ consecutive commented-out code lines (not doc comments)
    • Common placeholder strings: TODO, FIXME, HACK, XXX, impl later, stub, placeholder
    • Exclude: *.md, *.txt, documentation files
    • WARNING (non-blocking) for TODO/FIXME; BLOCKER for stub/placeholder in non-test files.
  4. Naming check — Use Read on each staged source file. Load naming conventions from CLAUDE.md if present (Grep for "naming" section). Apply:
    • Flag obvious violations: ALL_CAPS variable names outside constants, single-letter variables outside loop counters, names < 3 chars in function signatures
    • WARNING (non-blocking).
  5. No direct main check — Use Bash to confirm the current branch is NOT main:
    git branch --show-current
    If on main, BLOCKER: "You are committing directly to main. Create a branch first."
  6. Trailing whitespace / merge conflicts — Use Grep across staged files for:
    • <<<<<<< HEAD — unfixed merge conflict marker → BLOCKER
    • Trailing whitespace on lines → WARNING
  7. Report results — Print a summary table:
    Pre-Commit Gate Results
    ──────────────────────
    ✓ Secret scan       PASSED
    ✗ Dead code         BLOCKED — src/foo/handler.ts:42: "stub"
    ⚠ Naming            WARNING — 2 issues (non-blocking)
    ✓ Branch check      PASSED (branch: feature/my-feature)
    ✓ Conflict markers  PASSED
    
    Status: BLOCKED — fix 1 blocker before committing.
  8. If any BLOCKER exists: exit with a non-zero message and do NOT proceed with the commit. If only WARNINGs: inform the user and ask "Proceed with commit anyway? (y/n)" If all PASSED: confirm "All checks passed. Safe to commit."

Mode B — Install as git hook

When invoked as /pre-commit install:

Conventions shared with other skills that install a pre-commit hook (doc-refresh uses the same ones):

  • Marker: the hook's second line is the exact comment # claude-skill-hook: pre-commit. Install and uninstall only treat a hook as this skill's own when that line is present.
  • Backup and chain: an existing foreign hook is moved to pre-commit.backup.pre-commit in the same hooks directory, and the new hook runs it first, passing its arguments through and stopping on a non-zero exit. Each skill uses its own backup name (pre-commit.backup.<skill-name>), so two skills can stack without overwriting each other's backup.
  1. Use Bash to confirm this is a git work tree and resolve the hooks directory. --git-path hooks honors core.hooksPath and linked worktrees, unlike a hard-coded .git/hooks:
    git rev-parse --is-inside-work-tree
    HOOKS_DIR=$(git rev-parse --git-path hooks)
    git config --get core.hooksPath || true
    echo "$HOOKS_DIR"
  2. If core.hooksPath is set, a hook manager such as husky or lefthook probably owns the hooks. Warn the user that a file written to $HOOKS_DIR may be overwritten by that tool or committed to the repo, and suggest calling the checks from the manager's own config instead. Stop unless the user explicitly says to install anyway.
  3. Inspect $HOOKS_DIR/pre-commit:
    • Missing: continue to step 4.
    • Contains the line # claude-skill-hook: pre-commit: this is an earlier install of this skill. Overwrite it in place (an upgrade); do not create a backup.
    • Any other hook: if $HOOKS_DIR/pre-commit.backup.pre-commit already exists, stop and tell the user. Never overwrite an existing backup — it may be the only copy of their original hook. Otherwise move the hook aside, keeping its mode:
      mv "$HOOKS_DIR/pre-commit" "$HOOKS_DIR/pre-commit.backup.pre-commit"
      Tell the user the existing hook was backed up and will run first on every commit.
  4. Use mkdir -p "$HOOKS_DIR", then Write to create $HOOKS_DIR/pre-commit with the canonical script below verbatim — do not regenerate, paraphrase, or "improve" it. The deterministic checks are the enforcement path; the optional Claude call is advisory only and can never block or approve a commit.

    #!/usr/bin/env bash
    # claude-skill-hook: pre-commit
    # pre-commit hook installed by the pre-commit skill.
    # Enforcement is deterministic (grep-based). Claude, if present, is advisory only.
    # Bypass once: SKIP_PRE_COMMIT=1 git commit ...
    # (git commit --no-verify also works but skips ALL hooks, including this secret scan.)
    set -euo pipefail
    
    # Chain the hook that was here before install, if any. It runs even when
    # SKIP_PRE_COMMIT is set, because SKIP_PRE_COMMIT only skips this hook.
    CHAINED="$(dirname "$0")/pre-commit.backup.pre-commit"
    if [ -x "$CHAINED" ]; then
      "$CHAINED" "$@" || exit $?
    fi
    
    if [ -n "${SKIP_PRE_COMMIT:-}" ]; then
      exit 0
    fi
    
    STAGED=$(git diff --cached --name-only --diff-filter=ACMR || true)
    if [ -z "$STAGED" ]; then
      exit 0
    fi
    
    FAIL=0
    
    # 1. Block direct commits to main/master
    BRANCH=$(git branch --show-current)
    if [ "$BRANCH" = "main" ] || [ "$BRANCH" = "master" ]; then
      echo "BLOCKED: committing directly to $BRANCH. Create a branch first." >&2
      FAIL=1
    fi
    
    # 2. Secret scan on staged content (redacts matches in output).
    #    Test files are scanned too, but a match there is a warning, not a block.
    SECRET_RE='-----BEGIN (RSA|DSA|EC|OPENSSH) PRIVATE KEY-----|AKIA[0-9A-Z]{16}|sk-ant-[A-Za-z0-9_-]{10,}|sk-proj-[A-Za-z0-9_-]{10,}|github_pat_[A-Za-z0-9_]{20,}|ghp_[A-Za-z0-9]{30,}|glpat-[A-Za-z0-9_-]{20,}|npm_[A-Za-z0-9]{36}|AIza[0-9A-Za-z_-]{35}|xox[baprs]-[A-Za-z0-9-]{10,}|sk_live_[A-Za-z0-9]{16,}'
    ASSIGN_RE="(password|passwd|pwd|api_key|apikey|api_token|access_token|secret_key|client_secret|aws_secret_access_key)[[:space:]]*[=:][[:space:]]*['\"][^'\"]{8,}['\"]|export[[:space:]]+[A-Z_]*(KEY|TOKEN|SECRET|PASSWORD)[A-Z_]*=[^[:space:]]{8,}"
    
    while IFS= read -r f; do
      case "$f" in
        *.example|*.sample|node_modules/*|vendor/*) continue ;;
      esac
      MATCHES=$(git show ":$f" 2>/dev/null | grep -Ec -e "$SECRET_RE" || true)
      ASSIGNS=$(git show ":$f" 2>/dev/null | grep -icE -e "$ASSIGN_RE" || true)
      if [ "${MATCHES:-0}" -gt 0 ] || [ "${ASSIGNS:-0}" -gt 0 ]; then
        case "$f" in
          *.test.*|*.spec.*)
            echo "WARNING: possible secret in staged test file: $f (matches redacted; confirm it is a fake value)" >&2
            ;;
          *)
            echo "BLOCKED: possible secret in staged file: $f (matches redacted; inspect the file)" >&2
            FAIL=1
            ;;
        esac
      fi
    done <<< "$STAGED"
    
    # 3. Merge conflict markers
    while IFS= read -r f; do
      if git show ":$f" 2>/dev/null | grep -qE '^(<<<<<<< |>>>>>>> )'; then
        echo "BLOCKED: merge conflict markers in staged file: $f" >&2
        FAIL=1
      fi
    done <<< "$STAGED"
    
    if [ "$FAIL" -ne 0 ]; then
      echo "" >&2
      echo "pre-commit: commit blocked by deterministic checks above." >&2
      echo "Bypass once: SKIP_PRE_COMMIT=1 git commit ..." >&2
      exit 1
    fi
    
    # 4. Optional advisory review — output is informational only and never gates.
    #    PRE_COMMIT_ADVISORY guards against recursion (claude may run git commands).
    if [ -z "${PRE_COMMIT_ADVISORY:-}" ] && command -v claude >/dev/null 2>&1; then
      if command -v timeout >/dev/null 2>&1; then
        PRE_COMMIT_ADVISORY=1 SKIP_PRE_COMMIT=1 timeout 60 claude -p "/pre-commit" 2>/dev/null || true
      else
        PRE_COMMIT_ADVISORY=1 SKIP_PRE_COMMIT=1 claude -p "/pre-commit" 2>/dev/null || true
      fi
      echo "pre-commit: advisory review done (warnings above, if any, are non-blocking)."
    fi
    
    exit 0
  5. Use Bash to make the hook executable: chmod +x "$HOOKS_DIR/pre-commit"
  6. Confirm installation, naming the hooks directory used and any backup created. Remind the user:
    • The hook runs locally only — team members must run /pre-commit install themselves (or add it to project setup scripts).
    • Skip once with SKIP_PRE_COMMIT=1 git commit .... A chained earlier hook still runs.
    • git commit --no-verify also bypasses it, but that skips ALL hooks including the secret scan — prefer SKIP_PRE_COMMIT=1.
    • The Claude advisory step is optional, capped at 60 seconds where timeout exists, and can only warn — the deterministic checks are what block.

Mode C — Uninstall hook

When invoked as /pre-commit uninstall:

  1. Use Bash to resolve the hooks directory: HOOKS_DIR=$(git rev-parse --git-path hooks).
  2. If $HOOKS_DIR/pre-commit does not exist, report that no hook is installed and stop.
  3. Check for the marker: grep -qx '# claude-skill-hook: pre-commit' "$HOOKS_DIR/pre-commit". If it is absent, the active hook belongs to something else (another tool, or another skill installed on top of this one). Do not touch it; tell the user which hook is active and stop.
  4. If $HOOKS_DIR/pre-commit.backup.pre-commit exists, restore it as the active hook so the user's original hook keeps running:
    mv "$HOOKS_DIR/pre-commit.backup.pre-commit" "$HOOKS_DIR/pre-commit"
    Otherwise remove the hook: rm "$HOOKS_DIR/pre-commit".
  5. Confirm what was removed and whether a previous hook was restored.

Download now (ZIP)

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
What did you use it for? e.g. Reviewing pull requests in a Rails app

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.