The Agent Skills Directory

[COMMAND_EXECUTION]: The skill documentation includes instructions for executing local Python scripts used for developer optimization and evaluation cycles. Evidence: The file references/copilot_proposer_prompt.md describes a python -c script for keyword analysis, and references/program.md outlines a command-line workflow for running a routing accuracy evaluation script.
[PROMPT_INJECTION]: The skill reads from local files that may contain user-provided content, presenting a surface for indirect prompt injection. Ingestion points: exploration/session-brief.md and exploration/exploration-dashboard.md. Boundary markers: None. Capability inventory: Use of Read and Write tools alongside local command execution capabilities defined in reference files. Sanitization: No explicit sanitization of the input file content is performed before the agent processes it.

discovery-planning