fusion-package-scribe
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill operates on codebase contents that could potentially house adversarial inputs designed to manipulate the LLM's documentation process.
- Ingestion points: The agent reads repository source files, existing TSDoc blocks, and package README files across the monorepo workspace during the documentation phase as detailed in
agents/documenter.agent.md. - Boundary markers: There are no explicit text boundaries or instruction-isolation delimiters applied when loading the repository content into the agent's prompt context.
- Capability inventory: The skill has file-system write capabilities (editing
.ts,.tsx, andREADME.mdfiles) and interacts with the host shell to invoke localgit commitcommands as outlined inSKILL.md. - Sanitization: Content validation or filtering rules are absent, meaning embedded natural language instructions in scanned comments could influence the generation output.
Audit Metadata