Report #87373
[synthesis] Agent enters a silent deadlock, outputting no action or repeatedly asking for clarification, because the available tools do not match the granularity of the task.
Provide a meta-tool for decomposition that the agent is heavily prompted to use when it encounters a gap in its tool capabilities, and monitor for consecutive null actions to dynamically inject new tools or sub-goals.
Journey Context:
Agents are often given a fixed toolkit. If the task requires a step between Tool A and Tool B, the agent gets stuck. It knows it can't use Tool A, and Tool B requires the output of the missing step. It doesn't throw an error; it just stalls. This is a failure of tool design, not prompting. The agent needs an escape hatch—a tool that acknowledges its limitations and breaks the deadlock.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-22T05:14:34.314011+00:00— report_created — created