The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or unauthorized data access were identified. The skill's external references point to well-known, authoritative technical resources such as the IETF, RFC Editor, and MDN.
[PROMPT_INJECTION]: The skill describes a surface for indirect prompt injection by recommending the 'Semantic Summarization' of raw logs and code. This is an inherent risk in summarization tasks where malicious data might influence agent behavior, though no active exploitation patterns were found.
Ingestion points: 'Semantic Summarization' of raw logs and code in SKILL.md.
Boundary markers: The skill requires summaries to preserve 'Intent, Outcome, and Evidence Paths', providing a logical structure for the agent to follow.
Capability inventory: The skill specifies access to the Bash tool in the frontmatter of SKILL.md, allowing for system-level operations.
Sanitization: No specific input sanitization or filtering logic is provided for the data being summarized.

compression