The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is configured to execute a custom local runner (node scripts/security-audit.mjs) and a wide array of industry-standard security tools including npm audit, pip-audit, govulncheck, cargo audit, osv-scanner, semgrep, bandit, nuclei, and owasp zap to perform comprehensive security checks.
[PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection because its primary function involves the agent reading and triaging data from potentially untrusted sources.
Ingestion points: Local project files (source code, lockfiles, deployment configs), API specifications, and live content fetched from target URLs.
Boundary markers: The instructions do not specify the use of clear delimiters to isolate the ingested target data from the agent's core instructions.
Capability inventory: The skill has access to perform network requests, read/write to the local file system, and execute various shell commands.
Sanitization: The instructions do not define specific sanitization or filtering steps for the data retrieved from the target before it is processed for reporting.
[DATA_EXFILTRATION]: The skill performs network operations by fetching content from target URLs and running active security scanners. This behavior is necessary for its stated purpose of auditing websites. The skill explicitly mitigates the risk of unauthorized data exposure by instructing the agent to never print full secrets and to redact any sensitive values found during the audit.

security-audit