synthesis-agent-correspondence
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill facilitates an indirect prompt injection attack surface because it instructs the agent to ingest and process external, untrusted content from communication threads.
- Ingestion points: The
SKILL.mdfile defines a 'Reply-history gate' that requires the agent to 'read the entire thread, including the quoted history' from email and other channels before composing a response. - Boundary markers: The instructions do not define specific delimiters or 'ignore embedded instructions' markers to be used when interpolating ingested thread content into the agent's prompt context.
- Capability inventory: The agent is granted the capability to compose and send messages across Slack, email, and other outbound channels.
- Sanitization: While the skill references external voice and quality gates, it lacks explicit sanitization or filtering protocols for identifying or neutralising adversarial instructions embedded in the ingested thread history.
Audit Metadata