market-context
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 including
/foundation/brain.mdand/context/meta-patterns.md. Maliciously crafted content within these files could potentially influence the agent's logic during the market narrative construction. - Ingestion points:
/foundation/brain.md(Sections 3 and 5),/context/meta-patterns.md(guardrails). - Boundary markers: The instructions do not define explicit delimiters or escaping for the interpolated brain context during the prompt construction phases (Steps 0-3).
- Capability inventory: The skill possesses file-write capabilities, specifically updating Section 5 of
/foundation/brain.mdand appending to/context/skill-sessions.md. - Sanitization: A strong mitigation is present in Step 4, which mandates showing the user an exact before/after diff and requiring explicit confirmation before any write operation is executed.
Audit Metadata