The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/check_adr_gate.py utilizes subprocess.check_output to execute git diff for identifying changed files within the repository. The implementation is secure as it uses a list-based argument structure and avoids the shell, effectively mitigating shell command injection vulnerabilities. The script is intended for use in CI/CD environments to enforce project governance rules.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) by processing untrusted local content from markdown files and test source code. 1. Ingestion points: Content is read in scripts/build_adr_registry.py (ADR markdown), scripts/check_test_signal.py (test source code), and scripts/check_adr_gate.py (git output). 2. Boundary markers: Absent. The scripts provide structured JSON output to the agent but do not wrap extracted file content in explicit delimiters to prevent the agent from interpreting embedded instructions. 3. Capability inventory: Limited to local file system reads and git metadata operations via subprocess in scripts/check_adr_gate.py. 4. Sanitization: Absent. The scripts employ regex and JSON parsing to extract specific fields but do not filter the resulting text for natural language instructions.

ub-governance