wiki-integrate
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill performs filesystem read and write operations to maintain the wiki structure (SKILL.md). It implements a mandatory safety check (Config Discovery Step 2) that terminates execution if the agent's scope is set to a drive root, OS root, or user home directory. Additionally, it enforces a 'blacklist' check (Step 2) to ensure it does not modify unauthorized paths defined in the configuration.
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted content from user-created wiki pages to identify link candidates.
- Ingestion points: Reads full content of the target page (Step 1) and candidate pages (Step 5).
- Boundary markers: Explicit boundary markers are absent for the ingested page content.
- Capability inventory: Filesystem read and write access across the wiki root.
- Sanitization: The skill mitigates risks by imposing strict constraints in Step 6, which explicitly forbids the agent from rewriting or restructuring existing content, limiting edits to only adding wikilinks and updating frontmatter timestamps.
Audit Metadata