The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's verification test script (tests/run-tests.sh) clones the bats-core library from GitHub. This is a well-known, open-source testing framework used solely for validating the skill's utility scripts.
[COMMAND_EXECUTION]: The audit process executes several git commands (including git log, git grep, and git blame) and a local bash script to mine historical signals from the repository. These operations are essential for identifying stale files and hotspots and are restricted to the local filesystem.
[DATA_EXFILTRATION]: Although the skill performs comprehensive analysis on project source files and history, it includes no tools or instructions for transmitting data to external servers. All generated audit artifacts are stored locally within the .agents/tidy/project/ directory.
[PROMPT_INJECTION]: The skill processes untrusted repository content during its analysis phases, which introduces a surface for indirect prompt injection.
Ingestion points: The audit workflow reads all files within the repository to build inventories and analyze code areas.
Boundary markers: The sub-agent instructions do not currently utilize unique delimiters or explicit isolation warnings for the code content being processed.
Capability inventory: Sub-agents have permissions to write files within the audit directory and create additional sub-tasks for analysis.
Sanitization: Content from the repository is processed directly without sanitization or escaping of potential instruction-like strings.

tidy-project