The Agent Skills Directory

[COMMAND_EXECUTION]: The Preamble section contains a bash script that executes an external file (check-update.sh) located two levels above the skill's directory (../../check-update.sh). This execution of code outside the skill's own package is a risk if the environment's directory structure is not secured.
[DYNAMIC_EXECUTION]: In the 'Subagent Parallel Analysis' section, instructions for subagents describe a process to locate and run a Python script (anysearch_cli.py) by checking several hidden paths in the user's home directory (e.g., ~/.claude/, ~/.cursor/). Executing scripts from computed paths found during runtime discovery is a high-risk dynamic execution pattern.
[INDIRECT_PROMPT_INJECTION]: The skill collects qualitative data from the user via AskUserQuestion (including background information, constraints, and evaluation scores) and interpolates this content into a 'Strategic Decision Recommendation Report' and subagent prompts. The lack of boundary markers or sanitization for this external input allows for potential indirect prompt injection attacks where malicious data could influence the agent's logic or subagent behavior.

pm-decision