Gardening Skills Wiki
Pass
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: LOWSAFE
Full Analysis
- [INDIRECT_PROMPT_INJECTION] (INFO): The skill identifies an attack surface for indirect prompt injection because it reads and processes the content of markdown files. 1. Ingestion points: Scripts such as check-links.sh and check-naming.sh read SKILL.md and INDEX.md files. 2. Boundary markers: Absent. 3. Capability inventory: Display only; the scripts echo findings to the console for the agent. 4. Sanitization: Absent. The severity is INFO because the skill only provides diagnostics and cannot execute external instructions or communicate with the network.
- [COMMAND_EXECUTION] (LOW): The skill uses local bash scripts to run system utilities like find, grep, and sed for wiki maintenance. All execution is limited to local file analysis and diagnostic reporting within the skill's own environment.
Audit Metadata