run-hcom-agents
Pass
Audited by Gen Agent Trust Hub on Apr 21, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No malicious patterns or security vulnerabilities were identified in the skill's instructions, scripts, or configuration. The behavior aligns with the stated purpose of agent orchestration.
- [DATA_EXFILTRATION]: The skill documents and uses a relay mechanism for cross-device communication via well-known public MQTT brokers (broker.emqx.io, broker.hivemq.com, test.mosquitto.org). This is an optional, documented feature for distributed agent coordination.
- [INDIRECT_PROMPT_INJECTION]: The skill manages agent-to-agent communication, which represents an indirect prompt injection surface. The architecture includes mitigation measures such as XML-based context wrapping (<hcom_system_context>) and strict response protocols based on message intent (request, inform, ack). Mandatory evidence chain: 1. Ingestion points: hcom send and hcom transcript (all script files in references/scripts/); 2. Boundary markers: <hcom_system_context> tags (references/how-it-works.md); 3. Capability inventory: Subprocess execution via Claude Code and Codex agents (SKILL.md, references/cli-reference.md); 4. Sanitization: Intent-based routing and protocol-defined response rules.
Audit Metadata