firecrawl-core-workflow-a
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from external websites, creating a surface for potential indirect prompt injection.
- Ingestion points: Content retrieved from arbitrary URLs through
scrapeandcrawloperations. - Boundary markers: The instructions specify validation of metadata (e.g.,
statusCode,sourceURL) and required content fields to ensure data integrity. - Capability inventory: The skill is configured to use
WriteandEdittools for documenting changes or storing results. - Sanitization: The skill mandates writing only approved fields to downstream stores and quarantining unsuccessful content based on status codes.
- [EXTERNAL_DOWNLOADS]: The skill references official SDKs and documentation required for its stated purpose.
- References official Firecrawl documentation and GitHub repositories for implementation guidance.
- Mentions the
firecrawl(Node.js) andfirecrawl-py(Python) packages as the primary clients. - [CREDENTIALS_UNSAFE]: The skill manages an API key for Firecrawl Cloud operations.
- It includes strict instructions to inject the
FIRECRAWL_API_KEYfrom an approved secret manager and explicitly forbids leaking the key in source code, logs, or URLs.
Audit Metadata