The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted web content scraped via the FireCrawl API, which could potentially contain malicious instructions intended to influence the agent's behavior when it reads the generated markdown files.
Ingestion points: The scripts/firecrawl_research.py script fetches content from arbitrary URLs via the FireCrawl /v1/search endpoint.
Capability inventory: The skill has the ability to write to the file system and execute external document conversion tools.
Sanitization: Scraped content is truncated to 3000 characters, but no content-based filtering or sanitization is performed.
Boundary markers: The output files do not contain explicit delimiters or instructions to ignore embedded prompts.
[COMMAND_EXECUTION]: The scripts/convert_academic.py script invokes external binaries (pandoc and myst) to perform document conversion.
The execution is handled via subprocess.run using argument lists, which effectively prevents shell-based command injection.
The scope of execution is restricted to the specific tools required for the skill's stated academic research purpose.

firecrawl-research