Subagent spawn fails with 'No reasoning level selected' (reasoning-level-missing) — no default applied when caller omits reasoning level

Open
#15 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
45/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Active
Tech stack
python
Domain
backend, tooling

Research direction

Trace the subagent request path through the validation that emits reasoning-level-missing, then inspect how providers and models declare reasoning variants and their defaults. Confirm the behavior when the caller omits a reasoning level and make the chosen fallback behavior verifiable without reproducing the provider-specific account conditions.

Written by the indexing model from the issue text.

Description

Environment

  • OS: Windows 11 (build 10.0.26200)
  • ZCode: desktop client
  • Provider/account: zai coding plan, model GLM-5.3-Flash

Symptom

Spawning a subagent (Agent tool) fails with:

Cannot start subagent: No reasoning level selected
[reason=reasoning-level-missing]

Observed on 2026-09-13, 09-14, and 09-17 across different sessions on the same
account. Within an affected period it is persistent across retries — not
transient — and blocks all subagent use; the workaround is performing the work
directly in the main agent.

Analysis

The subagent request path validates the reasoning level before dispatch, and
when the caller did not supply one the validation fails with
reasoning-level-missing rather than falling back to anything. Providers and
models in this setup already declare reasoning variants with a default variant,
so a sensible default exists — it just isn't applied on this code path. It also
appears account/provider-selective (same client version worked at other times).

Suggested fix

Either:

  1. Fall back to the model's declared default reasoning variant when the caller
    supplies no reasoning level, or
  2. Inherit the current session's reasoning level for subagents, or
  3. Expose a setting (e.g. a default subagent reasoning level) so users can pin
    it.

Any of the three removes a hard blocker that currently makes subagent use
unreliable on this provider.

Dominant language
Python
Stars
39
Forks
20
No merged PRs in 30d

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from zai-org/zcode-plugins

All issues in zai-org/zcode-plugins

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.