Triad Derivations
Pass
Audited by Gen Agent Trust Hub on Apr 16, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data from local documentation files (VISION.md, SPEC.md, BUILDING.md) and interpolates that data into its reasoning process and outputs without security boundaries.
- Ingestion points: Content is read from
VISION.md,SPEC.md, andBUILDING.mdacross all derivation commands and automation triggers inSKILL.mdandAUTOMATION.md. - Boundary markers: Absent. The instructions do not define delimiters or provide the agent with guidance to ignore instructions that might be embedded within the processed documentation.
- Capability inventory: The skill has file-read access, the ability to prompt the user with structured questions (
AskUserQuestion), and instructions to suggest or execute document reconciliation commands (/reconcile). - Sanitization: Absent. The logic extracts and re-interpolates raw text such as 'Soul statements' and 'capabilities' into new narratives and reports, which could allow malicious instructions to persist across sessions.
Audit Metadata