The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the researchkit CLI via uv run to manage multi-turn sessions with various AI harnesses.
[EXTERNAL_DOWNLOADS]: The use of uv run researchkit may result in the automatic download and installation of the researchkit package from external registries if it is not already available.
[DATA_EXFILTRATION]: The skill accesses the ~/.researchkit/sessions/ directory for persistent storage and transmits repository data to external AI provider harnesses (e.g., Claude Code, Grok CLI) to generate project-aware advice.
[INDIRECT_PROMPT_INJECTION]: The skill processes repository data as untrusted input for AI agents, creating an attack surface for indirect prompt injection.
Ingestion points: The skill explicitly uses read, glob, and grep operations to ingest project files as context.
Boundary markers: No delimiters or explicit safety instructions for the agents regarding repository content were found.
Capability inventory: The skill can execute shell commands through the researchkit CLI wrapper.
Sanitization: No sanitization or input validation for ingested repository content is documented.

advise-session