The Agent Skills Directory

[COMMAND_EXECUTION]: The scripts explore_project.py and migrate.py utilize subprocess.run to interact with the local git binary. This functionality is used to analyze project history, detect uncommitted changes, and perform git mv operations to maintain file history during reorganizations. These operations are scoped to the project directory and are legitimate for the skill's stated purpose.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface because it reads and processes user-controlled project files and directory structures to generate summaries and status reports.
Ingestion points: Directory contents scanned by explore_project.py and specific project files (e.g., PROJECT.md, LOG.md) parsed by project_status.py.
Boundary markers: The tools produce structured text or JSON output, providing clear boundaries for the agent.
Capability inventory: The skill can move files (migrate.py) and create directories/templates (init_project.py). It lacks network access and arbitrary code execution capabilities.
Sanitization: migrate.py includes a resolve function that performs strict path validation to prevent path traversal attacks, ensuring file operations remain within the project root.
[SAFE]: No malicious patterns, unauthorized persistence, or exfiltration mechanisms were identified. The design explicitly includes safety features like the generation of undo scripts and mandatory reasoning for archiving files.

research-project