wiki-ingest
Pass
Audited by Gen Agent Trust Hub on May 15, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill implements a robust "Content Trust Boundary" for processing untrusted external data (PDFs, images, Markdown). It explicitly directs the agent to treat source content as information to be distilled rather than instructions to be executed.- [COMMAND_EXECUTION]: Employs local system utilities (
sha256sum,shasum) for file integrity and an optional CLI tool (qmd) for semantic search. The instructions emphasize secure shell patterns, including the use of--and double-quoting to prevent path-based command injection.- [PROMPT_INJECTION]: Includes specific guidance to ignore adversarial prompt patterns found in processed documents, such as "ignore previous instructions" or "disregard safety guidelines." This defensive instruction helps prevent the agent from being subverted by indirect prompt injection.
Audit Metadata