The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest data from several local files which may contain untrusted information from documents being reviewed. \n- Ingestion points: The skill reads output/CLAIMS.md, output/MISSING_EVIDENCE.md, output/NOVELTY_MATRIX.md, and DECISIONS.md to ground its generated review report. \n- Boundary markers: The prompt does not define boundary markers (like xml tags or unique delimiters) to separate the content of these input files from the skill's own instructions. \n- Capability inventory: The skill has no access to network tools, shell execution, or sensitive system credentials. Its actions are strictly limited to reading local files and writing a text report to output/REVIEW.md. \n- Sanitization: No input validation, escaping, or sanitization is performed on the data read from the local files before it is processed by the model.

rubric-writer