The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core function of processing untrusted documents.
Ingestion points: The skill accepts PDF, DOCX, and TXT files as input for contract analysis in SKILL.md.
Boundary markers: There are no explicit delimiters or specific warnings defined to isolate the document content from the AI's instruction set.
Capability inventory: The skill uses MCP tools including extract_text_from_pdf, extract_text_from_docx, and analyze_document_structure to ingest content.
Sanitization: No evidence of sanitization, filtering, or validation of the processed document content is present in the instructions.
[EXTERNAL_DOWNLOADS]: The skill references external resources hosted on GitHub.
Evidence: The README.md and SKILL.md files contain links to repositories and raw files within the claude-office-skills GitHub organization. These are recognized as legitimate vendor resources from the skill's author.

contract-review