The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data including user-generated content from reviews and support signals.
Ingestion points: references/feedback-learning-loop.md specifies the ingestion of "in-product free text", "customer-support cases", and "authorized public app-store, marketplace, and review-platform sources".
Boundary markers: While the skill requires preserving "original wording and source", there are no explicit boundary markers or instructions for the agent to ignore potentially malicious instructions embedded in the ingested feedback.
Capability inventory: No scripts, file-writing, or network-writing capabilities are included in the provided skill files.
Sanitization: The instructions do not define sanitization logic for filtering malicious prompt-like text within the feedback signals.
[EXTERNAL_DOWNLOADS]: The skill references external policy documentation from various platforms.
The skill provides links to official documentation for Google Play, Apple App Store, and Steam to guide the agent in verifying current platform policies.

run-product-feedback-loop