cso

Invoke the skill with a phrase like "security audit" or "OWASP review". It auto-detects your stack, then runs phased checks covering secrets, dependencies, CI/CD, webhooks, and LLM security, producing a severity-ranked report.

Daily mode applies an 8/10 confidence gate so only high-certainty findings are reported, keeping noise near zero. Comprehensive mode is a monthly deep scan with a 2/10 reporting bar that surfaces lower-confidence leads for manual review.

Yes. Phase 2 runs secrets archaeology using git log searches for known credential prefixes like AKIA, ghp_, sk-, and xoxb-, checks tracked .env files, and flags CI configs with inline credentials instead of secret stores.

Yes. Phase 4 inspects workflow files for unpinned third-party actions, dangerous pull_request_target triggers, script injection via github.event interpolation in run steps, and secrets exposed as environment variables.

Yes. Phase 7 greps for user input flowing into system prompts, unsanitized LLM output rendered as HTML, eval of AI responses, hardcoded AI API keys, and unvalidated tool or function calling.

Each phase ships explicit false-positive rules: placeholders and test fixtures are excluded, devDependency CVEs cap at MEDIUM, and user content in normal chat messages is not flagged as prompt injection. Findings require traced evidence before reporting.