The Agent Skills Directory

[SAFE]: The skill consists entirely of natural language instructions and organizational frameworks for product management tasks. No executable code or malicious logic was identified.
[NO_CODE]: The skill package does not include any scripts, binary files, or configuration for external tool execution.
[PROMPT_INJECTION]: The skill defines a role that processes untrusted data from sources like social media and customer reviews, which is an inherent surface for indirect prompt injection.
[INGESTION_POINTS]: Untrusted data enters the agent context via multi-channel collection (surveys, reviews, social media) as described in SKILL.md.
[BOUNDARY_MARKERS]: No explicit boundary markers or delimiters for untrusted content are defined in the instructions.
[CAPABILITY_INVENTORY]: The skill contains no executable capabilities, tool definitions, or file/network operations.
[SANITIZATION]: The 'Processing Pipeline' section specifies cleaning, normalization, and bias checking as manual or automated steps to validate data quality.

feedback-synthesizer