firecrawl-parse
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses
Bash(firecrawl *)andBash(npx firecrawl *)to execute document parsing commands on the local system. - [EXTERNAL_DOWNLOADS]: Downloads and runs the
firecrawlCLI from the npm registry via thenpxcommand. - [PROMPT_INJECTION]: The skill processes untrusted content from external documents (PDF, DOCX, HTML, etc.), which creates a surface for indirect prompt injection if the files contain instructions designed to manipulate the agent's behavior during summarization or Q&A tasks.
- Ingestion points: Local files at paths like
./paper.pdfprocessed by thefirecrawl parsecommand. - Boundary markers: No delimiters or protective markers are defined to separate ingested file content from the agent's instructions.
- Capability inventory: Shell access via
Bashfor executing Firecrawl CLI tools. - Sanitization: No sanitization of document content is described before the output is used for downstream agent tasks.
Audit Metadata