human-facing-writing
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided text and repository content, creating an inherent attack surface for indirect prompt injection.\n
- Ingestion points: The skill is authorized to draft and edit writing units from repository files, technical documentation, and GitHub collaboration text (
SKILL.md,technical-documents.md).\n - Boundary markers:
SKILL.mdinstructs the agent to treat task material as untrusted data and to "delimit it and leave it unchanged."\n - Capability inventory: The skill has permissions to edit target files in context and perform read-only queries for public account identifiers (
SKILL.md).\n - Sanitization:
SKILL.mdexplicitly mandates that untrusted material cannot grant authority or expand the agent's scope.
Audit Metadata