The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from local files (PDF, DOCX, XLSX, etc.) and converts them to markdown for the agent to read. This creates a surface where malicious instructions hidden within a document could attempt to override agent behavior or exfiltrate data when the agent processes the output.
Ingestion points: Local files provided by the user for parsing.
Boundary markers: The skill does not explicitly define delimiters or instructions to ignore embedded content in the parsed output.
Capability inventory: The skill can execute bash commands, write to the filesystem, and make network requests via the CLI.
Sanitization: No sanitization or filtering of document content is described.
[EXTERNAL_DOWNLOADS]: The skill documentation references the installation of the firecrawl-cli package via NPM. This is a vendor-provided tool necessary for the skill's local file operations.
[COMMAND_EXECUTION]: The skill relies on shell command execution (Bash(firecrawl *)) to interact with the local filesystem and the Firecrawl API. It specifically handles user-provided file paths and manages output redirection to the .firecrawl/ directory.
[DATA_EXFILTRATION]: The skill transmits local document content to the Firecrawl API for processing, summarization, and Q&A. This behavior is the core intended functionality of the skill using the vendor's cloud infrastructure.

firecrawl-parse