The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's workflow involves executing build commands and running related tests to verify the integrity of the project after restructuring. This is documented in SKILL.md under Workflow Step 5. The execution is bounded by user requests and the project context.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes arbitrary project files, which constitutes a surface for indirect prompt injection.
Ingestion points: Reads project manifests, entry points, and configuration files (e.g., package.json, .projectstructure.yaml) as specified in SKILL.md.
Boundary markers: None explicitly defined for project source data, although SKILL.md instructs the agent to treat example configurations as user settings rather than executable logic.
Capability inventory: The skill possesses the capability to move files, update imports, and run shell-based builds and tests.
Sanitization: The accompanying profile_store.py script performs sanitization on profile keys to prevent the storage of sensitive information, but the skill lacks specific sanitization for data found within the project files it reorganizes.
[SAFE]: The skill demonstrates several security best practices:
Credential Filtering: scripts/profile_store.py contains a SENSITIVE_PARTS block that prevents the persistence of keys containing terms like 'token', 'secret', 'api_key', etc.
Restrictive Permissions: The profile storage script enforces 0o600 (owner-only) permissions for the local JSON configuration file.
Safe Deserialization: scripts/validate_skill.py uses yaml.safe_load() to process configuration files, preventing arbitrary code execution during the validation phase.

lov-better-project-structure