The Agent Skills Directory

[SAFE]: No malicious patterns such as credential exfiltration, persistence mechanisms, or unauthorized privilege escalation were detected across the skill files.
[SAFE]: The codebase scanning and web search activities are strictly defined to gather information relevant to the user's specific problem space, which is consistent with the skill's stated purpose.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests data from external websites.
Ingestion points: External data enters the agent context via the WebFetch tool in SKILL.md (Step 3).
Boundary markers: The instructions include a specific mitigation requiring the agent to summarize content rather than reproducing it verbatim.
Capability inventory: The skill can perform web searches, fetch external content, and write files to the local .specs/ directory.
Sanitization: Content is sanitized through the summarization process, reducing the risk of executing embedded malicious instructions.

contextualize