apify-integration-development
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches documentation, quickstart guides, and OpenAPI schemas from official Apify domains (apify.com). These downloads are used for design planning and dynamic field generation.
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted web data from external Actors.
- Ingestion points: Data enters the agent context via Actor datasets and Key-Value store records (referenced in 'references/ai-harness-plugin.md').
- Boundary markers: Recommends implementing explicit delimiters (e.g., '<<<EXTERNAL_UNTRUSTED_CONTENT>>>') and source metadata to isolate untrusted content.
- Capability inventory: Involves API operations to start Actor runs, poll status, and fetch data items using the 'apify-client' SDK.
- Sanitization: Instructs on sanitizing forged delimiters and enforcing strict character limits (e.g., 50,000 characters) and item counts to prevent context hijacking.
- [DYNAMIC_EXECUTION]: Provides instructions for dynamically generating agent input models from remote JSON schemas. The guidance includes safety checks to ensure descriptions are pruned and validation constraints are correctly mapped to host platforms.
- [DATA_EXFILTRATION]: Documents proper handling of API tokens and OAuth2 credentials, advocating for the use of environment variables and secret managers while explicitly warning against hardcoding or logging secrets.
Audit Metadata