The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external files and codebases, which introduces a potential surface for indirect prompt injection from the target data. However, the skill explicitly includes mitigation instructions and best practices for agents to handle these risks.
Ingestion points: The target parameter in schemas/input.schema.json defines the files or directories to be read by the agent.
Boundary markers: The skill documentation in knowledge/owasp-agentic-ai-top-10.md (ASI08) instructs the agent to use clear delineation markers and "ignore instructions" headers for untrusted external content.
Capability inventory: The skill uses Read, Write, Edit, Bash, Glob, and Grep tools to conduct audits and generate reports.
Sanitization: The skill provides specific code patterns and instructional guardrails for validating and sanitizing retrieved content before it enters the context window.
[EXTERNAL_DOWNLOADS]: The skill manifest and documentation reference external tools and repositories for legitimate security research purposes.
The manifest.json file identifies semgrep and trufflehog as optional CLI dependencies, which are industry-standard security auditing tools.
It also includes research references to official OWASP GitHub repositories (CheatSheetSeries and ASVS) for security verification standards.
All referenced sources are well-known security organizations or established open-source projects.

security-architect