The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses npx to download and execute the @firecrawl/anydoc package (version 0.1.6) from the npm registry. This is a well-known service and the version is pinned for security.
[COMMAND_EXECUTION]: The Python wrapper in scripts/anydoc uses subprocess.run to invoke the npx command. The command is constructed safely using a list of arguments without shell=True, which prevents shell injection vulnerabilities.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection by processing untrusted document files and converting their contents into Markdown text for the AI agent.
Ingestion points: The convert and batch commands in scripts/anydoc receive document paths as input and process document contents.
Boundary markers: No explicit boundary markers or 'ignore' instructions for embedded content are added to the conversion output provided to the agent.
Capability inventory: The skill has the capability to run shell commands (via the Bash tool and its own npx wrapper) and read files (via the Read tool).
Sanitization: The conversion process does not sanitize or filter potential instructions embedded within the source documents.

anydoc