The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands like rm -rf and mkdir -p to manage internal project directories for temporary data storage.
[DATA_EXFILTRATION]: The skill performs a 'Project Deep Scan' that accesses sensitive files and metadata, including git commit history, source code comments (TODO/FIXME), and configuration files like package.json or docker-compose.yml which may contain infrastructure details.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its automated ingestion of untrusted external content.
Ingestion points: Processes external URLs provided in user input and scans project-level metadata such as Issue/PR descriptions and git history.
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are used when processing the collected data.
Capability inventory: The skill possesses the ability to modify the file system (via rm and mkdir) and orchestrate further downstream tasks like PRD generation and prototyping.
Sanitization: No sanitization or validation of the fetched URL content or project metadata is mentioned before it is processed by the refinement logic.

pm-need