best-practices-roundtable

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
  • [SAFE]: The skill is composed of best practices for model orchestration, emphasizing that model outputs are advisory and require deterministic local proof.
  • [COMMAND_EXECUTION]: The sanity.sh script uses grep and python3 to verify the structure and content of the skill's documentation.
  • [INDIRECT_PROMPT_INJECTION]: The skill manages data from multiple external model seats. Ingestion points: Model response files and join artifacts (SKILL.md). Boundary markers: Use of a structured Prompt Contract with defined objectives and constraints (SKILL.md). Capability inventory: Execution of the $ask tool and local proof commands like tests or database queries (SKILL.md). Sanitization: Explicitly treats model consensus as advisory and mandates local, deterministic verification before finalizing implementation work (SKILL.md).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 06:00 PM
Security Audit — agent-trust-hub — best-practices-roundtable