swarm-consensus
Pass
Audited by Gen Agent Trust Hub on Oct 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONNO_CODE
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill defines a coordination protocol where an agent processes natural language arguments from other agents during a 'Debate' phase. This creates an attack surface for indirect prompt injection, where a compromised agent could attempt to manipulate the consensus outcome or the behavior of the Coordinator.\n
- Ingestion points: Natural language arguments in the Debate phase (SKILL.md).\n
- Boundary markers: No specific delimiters or 'ignore' instructions are defined for the debate input.\n
- Capability inventory: The skill is purely instructional and does not include scripts or tool-use definitions.\n
- Sanitization: The skill includes high-level instructions for 'Red-flag-discard' of malformed outputs and 'Debate Hallucination Checks'.\n- [NO_CODE]: The skill consists solely of markdown instructions and YAML metadata. It does not include any executable scripts, binaries, or configuration files that perform automated actions.
Audit Metadata