The Agent Skills Directory

[SAFE]: The skill implements a structured security auditing workflow for project phases, focusing on verifying mitigations in documentation like SECURITY.md and PLAN.md.
[SAFE]: Explicit sensitive_data_rules are defined in the frontmatter to prevent the agent from accessing sensitive files, including .env*, secrets/**, *.pem, *.key, and credentials.json.
[SAFE]: The WebFetch tool is explicitly added to forbidden-tools, which effectively prevents any network-based data exfiltration attempts.
[SAFE]: Tool usage (Read, Write, Edit, Bash, Glob, Grep) is aligned with the skill's primary purpose of analyzing local project files and updating security documentation.
[SAFE]: The skill incorporates data redaction rules for output, specifically targeting tokens, credentials, and PII, which follows security best practices.

gsd-secure-phase