revenuecat-store-state
Pass
Audited by Gen Agent Trust Hub on Oct 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external store state data (prices, descriptions, availability) from third-party platforms like Apple and Google, creating a potential surface for indirect instructions. This is mitigated by robust procedural safeguards.
- Ingestion points: Data entering the agent context via the
get-product-store-statetool and CLI commands inSKILL.md. - Boundary markers: The skill explicitly mandates human intervention: 'Never call apply until you have shown the planned diffs and the user has explicitly confirmed.'
- Capability inventory: The skill uses
apply-product-store-state-planandsubmit-products-to-storeto perform writes to production environments. - Sanitization: Relies on user verification of summarized diffs and provides a dashboard URL for manual inspection of pending changes.
Audit Metadata