Repo security scan

Skill arkaigrowth/agent-skills/repo-security-scan

Nine production Claude Code skills and one command pack: repo security scanning, LLM-output validators, agent-fleet guardrails, credential hygiene. Deterministic cores, agentic edges.

Install

npx -y skills add arkaigrowth/agent-skills --skill repo-security-scan

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

2 things to look at

17 days oldThe repository was created 17 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Scan a repository for secrets, malware, risky dependencies, unsafe install scripts, and supply-chain issues. Use when evaluating whether a repo is safe to clone, run, or trust.

SKILL.md

5.2 KB, as published. Nobody here has run it

Repository Security Scan

Trigger: "scan repo", "security scan", "check repo security", "safe-clone", "quarantine clone", "audit this repo", "scan for vulnerabilities", "repo-scan", "is this repo safe"

Overview

Point-in-time security assessment for repositories using tiered scanning. Wraps widely used open-source tools: Trivy, TruffleHog, Gitleaks, GuardDog, OSSF Scorecard, Semgrep, Syft, and Grype. Each tool is optional: the scanner detects what is installed and skips the rest, so it still runs (with reduced coverage) when some tools are missing.

The scanner ships with this skill at scripts/repo-security-scan.sh. The commands below assume you run them from the skill directory; adjust the path if you invoke it from elsewhere.

Quick Commands

Scan current project

bash scripts/repo-security-scan.sh .

Scan with deeper analysis

bash scripts/repo-security-scan.sh --tier 2 /path/to/repo

Full deep scan

bash scripts/repo-security-scan.sh --tier 3 /path/to/repo

Quarantine clone (clone, scan, promote if clean)

bash scripts/repo-security-scan.sh --quarantine-clone https://github.com/user/repo ~/repo

JSON output (for automation)

bash scripts/repo-security-scan.sh --tier 1 --json /path/to/repo

Tiers

Tier	Approx time	Tools	Use when
1 (fast)	seconds	trivy, gitleaks, guarddog, heuristics	Every new clone, quick check
2 (standard)	1-3 min	+ scorecard, syft then grype, semgrep, trufflehog	Before using a repo for real work
3 (deep)	3+ min	+ full git history, license audit	Sensitive or high-stakes projects

Times are rough guidance and depend on repo size, installed tools, and network conditions. They are not guaranteed.

Quarantine Workflow

The safest approach for unknown repos:

Clone to quarantine: --quarantine-clone clones to ~/.quarantine/repos/ with --depth 1.
Auto-scan: Runs the specified tier of scans.
Verdict-based promotion:
- CLEAN (score >= 80): auto-promoted to the destination.
- WARN (score 50-79): promoted with a caution note.
- CRITICAL (score < 50): stays in quarantine, you decide.

Interpreting Results

Score and verdict

80-100 (CLEAN): no blocking findings. Review any info notes.
50-79 (WARN): warnings present. Review findings before proceeding.
0-49 (CRITICAL): do not run install scripts until findings are resolved.

Score starts at 100 and subtracts per finding: critical -25, high -10, medium -3, low -1 (floored at 0). Exit code follows the verdict: 0 for CLEAN, 1 for WARN, 2 for CRITICAL.

Severity

CRITICAL: live credentials, known malicious packages, active threats.
HIGH: unverified secrets, dangerous install scripts, .pth persistence files.
MEDIUM: outdated deps with known CVEs, missing lockfiles, npm install scripts.
LOW: minor vulns, style issues, info-level findings.

What Gets Scanned

Tier 1 (fast)

Trivy: vulnerabilities, secrets, misconfigurations in all files.
Gitleaks: secret/credential patterns in current files.
GuardDog: known malicious packages in Python/npm dependencies.
Heuristics: .pth files, dangerous setup.py calls, shell dotfiles, symlinks, .env files.
Posture: Dependabot, SECURITY.md, pre-commit hooks, CI security scanning.

Tier 2 (standard)

OSSF Scorecard: supply chain health score (requires GITHUB_TOKEN and a GitHub remote).
Syft then Grype: SBOM generation plus vulnerability scan.
Semgrep: static analysis with security-audit and secrets rulesets.
TruffleHog: deep secret scan with live credential verification.

Tier 3 (deep)

TruffleHog Git: full git history secret scan (every commit).
Trivy Licenses: license compliance audit.

Hook Integration

The skill bundles an optional PostToolUse hook at hooks/repo_security_hook.py. When wired into your Claude Code settings.json as a PostToolUse hook on Bash, it automatically triggers a Tier 1 scan when:

git clone is detected
npm ci or a project-level npm install is detected
pip install -r is detected

Other install commands (npm install <pkg>, pip install <pkg>, cargo add, brew install) trigger a reminder to scan.

The hook resolves the scanner relative to itself (../scripts/repo-security-scan.sh). Override the location with the REPO_SCAN_SCANNER environment variable if needed.

Reports

By default a markdown report is written to ./reports/repo-scan-{name}-{timestamp}.md. Override the directory with the REPORT_DIR environment variable, or pass --no-report to skip report generation.

Environment Variables

REPORT_DIR: report output directory (default ./reports).
QUARANTINE_DIR: quarantine clone directory (default ~/.quarantine/repos).
REPO_SCAN_TIMEOUT: per-tool timeout in seconds (default 60).
REPO_SCAN_LIB: optional path to a shell file defining send_alert for desktop notifications. Skipped if unset.
GITHUB_TOKEN: enables the OSSF Scorecard check in Tier 2.