site-visit-report

Pass

Audited by Gen Agent Trust Hub on Aug 14, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from field notes, external files, and user conversations. It mitigates potential injection risks by requiring rigid evidence classification (e.g., separating direct observations from participant-reported hearsay) and appending mandatory professional disclaimers to reports. Evidence chain: 1) Ingestion points: field-notes/*.md, photos, drawings, and conversation history. 2) Boundary markers: Enforced classification labels (O/R/I/ISS) and mandatory professional disclaimers. 3) Capability inventory: Bash, Read, Write, Glob, and Grep tools. 4) Sanitization: None specified.
  • [COMMAND_EXECUTION]: The skill utilizes the Bash tool for project context resolution and report file management. It follows safety best practices such as deterministic collision avoidance and project-relative path resolution, which are consistent with its stated administrative purpose.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 14, 2026, 03:45 PM
Security Audit — agent-trust-hub — site-visit-report