gardening-skills-wiki

Pass

Audited by Gen Agent Trust Hub on Sep 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill analyzes the content of local markdown files and search logs, which represents a surface for indirect prompt injection.
  • Ingestion points: check-links.sh, check-naming.sh, check-index-coverage.sh, and analyze-search-gaps.sh read all .md and .jsonl files in the target directory.
  • Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands within the files being processed.
  • Capability inventory: The scripts are restricted to read, search, and report operations using standard CLI tools (grep, sed, jq).
  • Sanitization: Log processing utilizes jq, and markdown parsing relies on specific regex patterns.
  • [COMMAND_EXECUTION]: The skill provides and executes local Bash scripts to perform maintenance tasks.
  • The implementation uses standard tools like find, grep, and sed to traverse and analyze the directory structure.
  • The garden.sh script manages permissions for the other included scripts using chmod.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 9, 2026, 03:52 AM
Security Audit — agent-trust-hub — gardening-skills-wiki