The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructs sub-agents to search for a script named anysearch_cli.py across several hidden directories, including ~/.claude/skills/anysearch/scripts/, ~/.cursor/skills/anysearch/scripts/, and others. If found, the agent executes the script using python3 via the Bash tool. This pattern of dynamically discovering and running external executables from computed paths poses a risk as the integrity and source of the script cannot be verified and it may originate from a different agent's installation.
[COMMAND_EXECUTION]: The skill performs shell command execution using the Bash tool to run the dynamically located anysearch_cli.py script with user-derived search queries.
[INDIRECT_PROMPT_INJECTION]: The skill reads existing product reports to guide the brainstorming process, creating a vulnerability to malicious instructions embedded in those documents.
Ingestion points: Reads from docs/01-需求调研/需求调研报告.md.
Boundary markers: None identified; the content is used directly in the logic.
Capability inventory: The skill has access to Bash, Write, Agent, and WebSearch tools.
Sanitization: No validation or sanitization is performed on the ingested content.

pm-brainstorm