skills/b-open-io/prompts/code-audit-scripts

code-audit-scripts

SKILL.md

Code Audit Scripts

Deterministic security and quality scans that output structured JSON. No LLM reasoning needed for the scanning — your job is to interpret results and recommend fixes.

Quick Start

Run everything at once:

bash <skill-path>/scripts/parallel-audit.sh /path/to/project

Returns a merged JSON report with all findings categorized by type and severity.

Individual Scans

Scan for Hardcoded Secrets

bash <skill-path>/scripts/scan-secrets.sh /path/to/project

Detects: API_KEY, SECRET, PASSWORD, PRIVATE_KEY, ACCESS_KEY, DATABASE_URL, JWT_SECRET, STRIPE_SK, and more. Filters out references to env vars (process.env, os.environ) to reduce false positives.

Scan for Debug Artifacts

bash <skill-path>/scripts/scan-debug.sh /path/to/project
# Include test files:
bash <skill-path>/scripts/scan-debug.sh /path/to/project --include-tests

Detects: console.log/debug/warn, debugger statements (JS/TS), print/breakpoint (Python), fmt.Println (Go). Skips test files by default.

Scan for TODOs and FIXMEs

bash <skill-path>/scripts/scan-todos.sh /path/to/project

Categorizes by severity:

  • High: FIXME, BUG, HACK, XXX — these need attention before shipping
  • Low: TODO — tracked work items

Acting on Results

Finding Type What to Do
Secrets with real values Immediately flag to user. Rotate the credential. Move to env var.
Secrets that are env var refs False positive — ignore
Debug artifacts in src/ Remove before shipping. List specific files and lines.
Debug artifacts in tests Usually fine. Only flag if excessive.
FIXME/HACK/XXX Flag as blockers for the current PR/deployment
TODO Informational. Mention count but don't block on them.

The parallel-audit.sh output includes a summary object with counts per category and high_priority count — use this for quick pass/fail decisions.

Weekly Installs
3
GitHub Stars
8
First Seen
6 days ago
Installed on
amp3
cline3
opencode3
cursor3
kimi-cli3
codex3