manage-roadmap
Pass
Audited by Gen Agent Trust Hub on Jun 29, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill operates entirely through instructional logic and standard agentic tools (TeamCreate, Agent). It does not include executable scripts, external binary dependencies, or unauthorized network operations.
- [PROMPT_INJECTION]: The skill uses directives such as 'Execute these instructions directly — do NOT delegate' and 'Do NOT analyze yourself'. These are functional orchestration constraints designed to maintain the Conclave framework's multi-agent logic and do not represent attempts to bypass safety filters.
- [DATA_EXFILTRATION]: No data exfiltration patterns were detected. The skill's 'Shared Principles' explicitly prohibit the inclusion of credentials, API keys, or PII in agent context or outputs, and it provides a 'no secrets in context' rule for developers.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests data from 'docs/ideas/' and 'docs/research/'. This represents a standard surface for indirect prompt injection common to analysis tasks. The risk is mitigated by the 'Lead-as-Skeptic' review gate, which requires a human-equivalent review of all rationale before roadmap changes are finalized.
Audit Metadata