The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill instructs sub-agents to locate a script named anysearch_cli.py within multiple hidden application directories (such as ~/.claude/, ~/.cursor/, and ~/.anysearch/) and execute it using python3. This behavior relies on dynamically determined file paths for code execution, which is a security risk if the environment is compromised or if malicious files are placed in these search paths.
[COMMAND_EXECUTION]: The skill employs bash to perform file system checks (e.g., verifying the existence of PRD documents) and to carry out the search-and-execute logic for the anysearch_cli.py utility.
[INDIRECT_PROMPT_INJECTION]: The skill processes a local PRD产品需求文档.md file and user inputs to generate business strategies. It lacks explicit boundary markers or sanitization to prevent malicious instructions within the PRD from influencing the agent's actions.
Ingestion points: Processes docs/02-方案设计/PRD产品需求文档.md and user-provided answers from AskUserQuestion.
Boundary markers: Absent; the content of the PRD is ingested directly into the prompt context without delimiters.
Capability inventory: The skill possesses Bash, Write, and Agent (sub-agent spawning) capabilities.
Sanitization: Absent; there is no evidence of validation or filtering for the ingested document content.

pm-business-model