The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes several diagnostic commands through a shell script to assess the local security posture.
Evidence in scripts/scan.sh includes lsof -nP -iTCP to detect listening ports, git ls-files and git log to find committed secrets, and npm audit for dependency checks.
The commands are scoped to the user's home directory and local environment, and are consistent with the skill's stated purpose of auditing exposure.
[CREDENTIALS_UNSAFE]: The skill actively searches for hardcoded credentials and secret patterns (API keys, private keys, .env files) across the file system and git history.
Evidence in scripts/scan.sh shows a comprehensive KEYPAT regex designed to detect common API key formats (Anthropic, GitHub, AWS, Slack, etc.).
Safety Mitigation: The skill explicitly implements a redaction rule in SKILL.md and scripts/scan.sh, ensuring that only the variable name and the first 7 characters of a key are ever printed or included in reports to prevent further exposure during the audit process.
[EXTERNAL_DOWNLOADS]: The skill uses curl to probe specific domains provided by the user.
Evidence in scripts/scan.sh shows curl is used with the -sI flag to check security headers and search for key material in public HTML bundles.
These requests are read-only and follow a strict "no-redirect" (no -L) policy to ensure accurate status reporting.

exposure-audit