message-house
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources and user input to generate messaging tables, which creates a potential surface for indirect prompt injection.
- Ingestion points: The skill loads content from
/foundation/brain.md(Sections 1–6) and accepts user-provided text or pasted output from other skills (Step 0, Steps 3–4). - Boundary markers: Absent. The instructions specify table formatting rules but do not implement explicit delimiters or "ignore embedded instructions" warnings for the content being interpolated.
- Capability inventory: The skill appends metadata to
/context/skill-sessions.md(Step 6) and outputs formatted markdown to the user. - Sanitization: Absent. The skill reformats source content (trimming length, bolding features) but does not filter or escape instructions embedded within the text.
Audit Metadata