The Agent Skills Directory

[SAFE]: The skill implements strict data integrity rules (Axioms) that require the AI to provide literal quotes from the source text as evidence for its conclusions. This effectively mitigates data hallucination and ensures traceability.
[EXTERNAL_DOWNLOADS]: The project includes a package.json file that specifies a dependency on ajv, a widely used and trusted JSON Schema validator. The instructions guide the user to install this package from the official NPM registry (npmjs.org), which is a well-known and secure service.
[COMMAND_EXECUTION]: The skill provides a validation script (scripts/validate.js) intended for local use to verify the AI's output. The script performs schema validation and cross-field integrity checks (e.g., ensuring referenced IDs exist). It does not perform network operations or access sensitive system files.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes untrusted user-supplied scripts (source_script). However, the requirement for literal substring matches and the structured output format serve as strong boundary markers. The documentation explicitly acknowledges semantic risks (e.g., sarcasm or refutation) and provides a heuristic 'hedge marker' check to identify potential issues.

100x-persona