The Agent Skills Directory

[COMMAND_EXECUTION]: The script evals/run_real_eval.py uses subprocess.Popen to execute the claude CLI. This is a functional testing utility designed to automate the evaluation of the skill's trigger rates by passing specific queries from local JSON files to the agent and monitoring tool output.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection due to its interaction with repository-level files.
Ingestion points: The agent is instructed to read CLAUDE.md for build/lint commands and ${CLAUDE_PLUGIN_ROOT}/references/change-type-labels.md for commit classification logic.
Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded commands within these external files.
Capability inventory: The skill enables the agent to perform Git operations (staging, committing, branch creation) and invoke the perform-preflight quality gate skill.
Sanitization: There is no evidence of sanitization or validation of the contents of the external configuration files before they are used to influence agent behavior.

committing-changes