The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a custom Python script, scripts/git_term_index.py, which executes local git commands via the subprocess module. These calls are constructed as argument lists and do not invoke a shell, protecting against command injection. The operations are restricted to non-destructive analysis of local repository history.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to analyze external project data, such as thesaurus files and commit logs, which are potentially untrusted ingestion points.\n
Ingestion points: The agent reads the project's THESAURUS.md file and uses a Python script to extract metadata from git log output.\n
Boundary markers: The skill specifies a rigid "grep-first" format for the thesaurus, using specific line patterns (Index, Terms, Forbidden, etc.) and tokens (kind:, avoid:, ctx:) to structure how the agent interprets the data.\n
Capability inventory: The skill is permitted to use Read, Write, Edit, Grep, Glob, and Bash tools.\n
Sanitization: The instructions guide the agent to act only upon the structural "shape" of the data hits, providing a layer of validation against instructions embedded in prose or unrelated comments.

ubiquitous-language