skills/c0nant/skills/setup-solid/Gen Agent Trust Hub

setup-solid

Pass

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: SAFE
Full Analysis
  • [DATA_EXPOSURE]: The skill reads local repository files such as CLAUDE.md, CONTEXT.md, and Architectural Decision Records (ADRs) to extract architectural patterns. This data is used solely to generate the documentation section and is not transmitted externally.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests potentially untrusted data from repository files (ADRs, existing documentation) and incorporates it into a high-authority instruction file (CLAUDE.md).
  • Ingestion points: Reads CLAUDE.md, AGENTS.md, CONTEXT.md, and .agents/adr/ from the repository root.
  • Boundary markers: The skill does not explicitly use escaping or boundary markers for the ingested text.
  • Capability inventory: Uses file reading and file writing tools to update project documentation.
  • Sanitization: The process includes a mandatory human-in-the-loop 'Draft and confirm' step (Step 2), where the agent must show the user the intended changes before writing, effectively mitigating the risk of malicious instructions being silently injected.
  • [SAFE]: The skill follows least-privilege principles by scope-limiting the documentation to the 'boy scout rule' and updating existing sections in place rather than overwriting files entirely. It correctly handles symlinks and monorepo configurations without escalating permissions.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 21, 2026, 01:08 PM
Security Audit — agent-trust-hub — setup-solid