The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local CommonJS script (evaluate-prompts.cjs) using the node environment via the Bash tool to generate usage metrics from internal logs. The script is invoked using an environment-relative path (CLAUDE_PLUGIN_ROOT).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data, specifically prompt history logs and user-supplied strings, as its primary input for evaluation.
Ingestion points: Analyzes logs at .claude/metrics/prompts/{date}.jsonl and user-provided content during optimization tasks.
Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions to prevent the agent from following directives contained within the data being evaluated.
Capability inventory: The skill uses Read, Bash, Glob, and Grep tools.
Sanitization: There is no evidence of content sanitization or escaping performed on the external prompt data before it is processed by the agent.

prompt-evaluator