The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external, untrusted codebases for security vulnerabilities. This creates an indirect prompt injection surface where an adversary could embed malicious instructions in code comments or data within a target project to influence the agent's behavior.
Ingestion points: Analysis of project source code, configuration files, and dependency manifests as described in the reconnaissance phase.
Capability inventory: The skill utilizes shell-based search tools (grep), static analysis engines (Semgrep, Bandit, Slither), and filesystem access to audit code.
Sanitization: The methodology emphasizes manual verification and proof-of-concept development, which acts as a human-in-the-loop filter, though it does not specify automated input sanitization for the agent's internal processing.
[SAFE]: The skill serves a purely defensive and educational purpose, providing high-quality resources for security professionals. All code examples (e.g., SQL injection, reentrancy) are clearly contextualized as "bad" versus "good" patterns. No unauthorized network operations, credential exfiltration, or persistence mechanisms were detected. The use of standard security tools like grep and Semgrep is consistent with the skill's stated intent.

vulnhunter