The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs file system modifications using shell commands in its optional apply-safe mode. Evidence: Stage 6 instructions describe using git mv for moving or renaming files and performing directory deletion for empty folders. These operations are restricted to safe, reversible actions and require explicit user confirmation.
[DATA_EXFILTRATION]: The skill processes repository structure and metadata, which creates a data exposure surface in the generated report. Evidence: Stage 1 and Stage 2 involve recursive scanning of the project directory and reading file contents to infer artifact types. Note: No network transmission occurs, and findings are limited to a local report file.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection from the untrusted repository data it processes. Ingestion points: File system enumeration in Stage 1 and content inspection in Stage 2. Boundary markers: Not explicitly defined for the ingested file data. Capability inventory: Modification of the local file system (move, rename, delete empty directories) during apply-safe execution. Sanitization: The instructions do not specify sanitization or validation of the ingested file data before processing.

tidy-repo