Persistent Agenda Management for Asynchronous Colleague Clarification
A system that maintains a persistent asynchronous dependency queue to track deferred colleague queries and automatically schedule follow-up messages before task completion.
Time
Compute level
Resources
What the paper used
Prerequisites
Scope
The Gap
Section 6.3.2 and Figure 8b analyze colleague clarification over time, tracking how agents handle deferred answers from digital employees across 450 deferred owner-card threads.
“The remaining 153 threads receive no follow-up (Figure 8b). At the initial deferral, these abandoned threads have a median of 7.1 hours of execution budget remaining, and the agent continues working for a median of 3.4 hours afterward.”
Limitation
Agents frequently abandon conversational threads when colleagues defer answers or ask for reproduction context, dropping 34% (153 of 450) of deferred threads despite having hours of execution budget remaining.
Proposed Solution
Construct an explicit stateful dialogue agenda within the agent scaffolding that registers every outgoing inquiry as an active promise. When a colleague defers or requests additional context, the agenda transitions the query to a 'waiting' state, binds it to pending code investigations, and triggers automated follow-up events whenever new execution traces become available or before any terminal handoff action can be invoked.
The paper’s authors list this direction as future work, so they may already be pursuing it.
Why This Matters
Closing abandoned clarification threads could recover a significant portion of the 42% of test failures currently caused by unacquired requirements.
How to Get Started
How to test it: Evaluate the stateful agenda scaffold against the baseline Claude Code scaffold across the 101 SWE-CoWork tasks; test whether the deferred thread abandonment rate drops below 10% and F2P pass rate improves by at least 5 percentage points, stopping if dialogue looping degrades overall task budget without improving disclosure.
- Clone the public repository at https://github.com/TokenWave-AI/swe-cowork-bench and configure the Docker evaluation container.
- Implement an interceptor in the agent scaffold that parses colleague responses for deferral tokens or pending conditions and appends them to a local JSON tracking queue.
- Add a pre-submission verification hook that rejects final handoff if any queue items remain in a deferred state without subsequent follow-up.
- Run evaluation on a 15-task subset of SWE-CoWork to measure follow-through rate and token overhead.
Suggested by AI. Adapt it to your setup.
Has This Been Done?
While papers such as 'Effective Strategies for Asynchronous Software Engineering Agents' and 'LLMs are General Asynchronous Agents' touch upon the broad domain of asynchronous agent coordination, they do not describe the specific mechanism of an explicit, stateful agenda that registers inquiries as 'promises' and forces wait-state transitions tied to pending execution traces. These existing works focus on high-level architecture rather than the fine-grained tracking of conversational dependencies.
- LLMs are General Asynchronous Agents · 2026 · 0 citations
- Effective Strategies for Asynchronous Software Engineering Agents · 2026 · 23 citations
We searched OpenAlex and arXiv for “agent dialogue state tracking asynchronous task follow-up” and “Persistent Agenda Management for Asynchronous Colleague Clarification” and an AI compared the idea with the top 10 results. A small search can miss relevant work.
Source Paper
SWE-CoWork: Benchmarking Coding Agents as Digital Employees in OrganizationsXinping Lei, Minghao Liu, Ruijie Zhu +23 more · 2026
The paper introduces SWE-CoWork, a benchmark evaluating software engineering agents as digital employees within simulated organizations of 26 rule-governed colleagues across 101 repositories, 11 languages, and 12 domains. Agents must inspect evolving records, converse with colleagues to recover distributed requirements, implement complex changes, and deliver a verified handoff. The study reveals a critical gap between partial code generation (~63% fail-to-pass tests) and end-to-end task success (<8%), primarily caused by breakdowns in long-horizon requirement tracking, abandoned clarifications, and hallucinated handoff completion.
Related Ideas
- Optimistic Synchronization for Cyclic DependenciesVery Large-Scale Multi-Agent Simulation in AgentScope🔴Hard
- Asynchronous RL for Multi-Turn Agentic TasksAReaL: A Large-Scale Asynchronous Reinforcement Learning System for Language Reasoning🟡Fairly easy
- Requirement-Coverage and Execution-Trace-Aware Candidate Deliverable SelectionHarness^2: Recursive Agent Harnessing for an Open World🟢Easy
- Proactive User-in-the-Loop Clarification for High-Ambiguity Daily TasksStep-GUI Technical Report🟠Medium
- Evaluating Continual Agent Adaptation on Open-World Human-in-the-Loop WorkloadsMetaClaw: Just Talk -- An Agent That Meta-Learns and Evolves in the Wild⛔Very hard