The Agent Skills Directory

[Indirect Prompt Injection] (MEDIUM): The skill's primary purpose is to ingest and analyze untrusted data from legacy codebases, creating a surface for indirect prompt injection.
Ingestion points: Reads all files within the user-specified directory via codecompass batch:index <path> and codecompass requirements:extract.
Boundary markers: Absent; the workflow does not specify delimiters or instructions to ignore embedded commands within the code being analyzed.
Capability inventory: The skill possesses the capability to execute shell commands and write files to the local system.
Sanitization: There is no evidence of sanitization or filtering to prevent the agent from obeying instructions hidden in code comments or string literals.
[Command Execution] (MEDIUM): The skill requires the execution of a specialized CLI tool (codecompass) that is not part of standard system utilities or a verified package from a trusted organization. This introduces risk if the binary has been tampered with or is malicious.
[Metadata Poisoning] (LOW): Automated scans flagged requirements.md as a malicious URL. This appears to be a false positive where a standard markdown output filename was misinterpreted as a blacklisted URL string, though it highlights the potential for deceptive naming in tool outputs.

extract-requirements