The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze untrusted external data, including app metadata from store APIs and manual text pastes. This creates a surface for indirect prompt injection where malicious instructions embedded in app names or descriptions could attempt to manipulate the agent's logic. \n * Ingestion points: External data retrieved via list_keywords, aso_search_results, and get_app_metadata, plus manual user metadata pastes. \n * Boundary markers: The instructions do not define clear delimiters or guardrail instructions to isolate untrusted content. \n * Capability inventory: The skill can modify state via the edit_keywords tool (following a user-confirmed loop) and writes HTML/JSON reports to the local file system. \n * Sanitization: No explicit sanitization of ingested app metadata is mentioned before interpolation into reports. \n- [EXTERNAL_DOWNLOADS]: The skill references documentation and support resources from official AppFollow domains (appfollow.io). These are vendor-owned resources and are used to provide informational context for ASO metrics.

aso-analysis