The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is primarily used to audit external application code and configuration files. This data ingestion path creates a surface for indirect prompt injection if an audited file contains malicious instructions designed to subvert the agent's logic during the audit.
Ingestion points: Uses Read, Grep, and Glob tools to retrieve content from project files as seen in SKILL.md and the OWASP reference.
Boundary markers: No specific delimiters or boundary markers are instructed for use when processing the audited content to distinguish it from system instructions.
Capability inventory: The skill utilizes the Bash and Agent tools, allowing for command execution and sub-agent dispatch (e.g., the yokay-security-scanner).
Sanitization: The instructions do not define methods for sanitizing audited code before it is interpreted or summarized by the agent.
[EXTERNAL_DOWNLOADS]: The references/dependency-security.md file documents the use of standard package managers to install well-known security auditing tools, including snyk, pip-audit, trivy, and syft. These are industry-standard tools from established providers.
[DATA_EXFILTRATION]: The authentication reference includes a utility that fetches data from api.pwnedpasswords.com, which is a well-known and reputable service used to check for compromised credentials. This is documented as a legitimate part of the auditing process.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform complex automated scans and runs security commands within the local environment, such as npm audit, gitleaks, and specialized grep patterns for vulnerability detection.

security-audit