codex-async-followup
Pass
Audited by Gen Agent Trust Hub on Sep 19, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill instructions define a standard workflow for managing long-running asynchronous tasks within the environment. It guides the agent on how to use the
automation_updatetool to schedule one-time heartbeats when waiting for background tasks. - [INDIRECT_PROMPT_INJECTION]: The skill facilitates the agent resuming work from a conversation thread after a delay, which involves ingesting data from the results of background tasks or subagents. This creates a surface for indirect prompt injection, but it is the intended primary purpose of the skill's asynchronous workflow management.
- Ingestion points: Conversation thread context following an automated heartbeat (SKILL.md).
- Boundary markers: None explicitly defined.
- Capability inventory: Relies on
automation_updatefor task scheduling. - Sanitization: No explicit sanitization of background task output is described.
Audit Metadata