write-hugo-twin-content
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external research data, previous posts, and stakeholder feedback to generate markdown files, which creates a surface for indirect prompt injection.
- Ingestion points: Untrusted data enters the context through retrieved research and prior posts (referenced in guardrail ASI06) and stakeholder input synthesized from other skills (referenced in Related Skills: meeting-review).
- Boundary markers: The skill provides structural instructions and a "Security Guardrails" section (ASI01-09) that explicitly directs the agent to verify claims and cross-check against source briefs.
- Capability inventory: The skill uses the
write_file,edit_file, andcreate_filetools to modify the workspace based on processed inputs. - Sanitization: Guardrail ASI09 (Human-Agent Trust Exploitation) specifically requires human editorial sign-off and disclosure of AI involvement to prevent unverified content publication.
Audit Metadata