The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's preamble automatically executes a shell script named check-update.sh from its own directory whenever the skill is loaded.
[REMOTE_CODE_EXECUTION]: The skill contains instructions for sub-agents to perform dynamic code execution. It mandates searching for a specific Python script, anysearch_cli.py, across multiple hidden application directories in the user's home folder (including ~/.claude/, ~/.cursor/, and ~/.anysearch/) and executing it using the Bash tool. Running executables discovered via dynamic path searching in sensitive system folders is a high-risk behavior.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because user-supplied input is interpolated directly into sub-agent prompts without sanitization or protective boundary markers.
Ingestion points: Product details and requirements collected via AskUserQuestion in steps 1 through 3 of the execution flow.
Boundary markers: Absent; user input is placed directly into the task descriptions for sub-agents.
Capability inventory: The skill uses the Agent tool to invoke sub-agents which have access to the Bash and WebSearch tools.
Sanitization: None; the agent does not escape or validate user-provided strings before using them in sub-agent prompts.

pm-demand