write-a-wiki-page

Pass

Audited by Gen Agent Trust Hub on Nov 17, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute project maintenance and build commands using pnpm, such as "pnpm check", "pnpm generate", and "pnpm build", which are standard for the defined development environment.- [EXTERNAL_DOWNLOADS]: The skill uses "pnpm install" for managing dependencies and "pnpm sources:fetch" to retrieve remote Git history for code snippet verification, targeting the project's own infrastructure.- [INDIRECT_PROMPT_INJECTION]: The skill processes content from local directories to generate documentation, representing an ingestion surface for untrusted data. Evidence Chain: (1) Ingestion: Local markdown files in .brain/ and skills/ directories. (2) Boundary markers: Markdown code fences with strict metadata keys (repo, path, at, lines). (3) Capabilities: Shell execution via pnpm and file modification in apps/site/scripts/. (4) Sanitization: The build process includes extensive diagnostic checks (e.g., CodeInvalidProps, CodeUnknownKey) to validate all content and opt-outs.
  • [DYNAMIC_EXECUTION]: TypeScript code snippets within documentation are extracted and type-checked during the build process, which involves generating temporary modules for compiler validation.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 17, 2026, 09:54 PM