fix(agents): restore global subagent model default priority over agent own model by joeykrug · Pull Request #58823 · openclaw/openclaw

joeykrug · 2026-04-01T07:25:23Z

Summary

Current exact-head refresh

True-merged the shared upstream cutoff 262baedcf7fa3d66dc35cce7d0253e0a31f0cc0e into exact branch head 49b4d13321fcda6e3ca3ef2744f17a0dca52749a.
The preceding refreshed head was green across CI, CodeQL, CodeQL Critical Quality, OpenGrep, Workflow Sanity, and iOS Periphery; fresh Actions are running on this final merge-only synchronization.
The live resolver proof below was run on the earlier patched head and is retained as prior real-behavior evidence; it was not re-run locally after the merge-only refresh.

Restores correct subagent model precedence in resolveSubagentConfiguredModelSelection: global agents.defaults.subagents.model now takes priority over the parent agent's own model.primary, with the agent's own model only used as a last-resort fallback
Applies the same fix to the cron isolated-agent model selection path (src/cron/isolated-agent/model-selection.ts) and the shared resolveSubagentModelConfigSelectionResult helper in src/agents/agent-scope.ts
Updates existing tests from Agents: fix subagent model precedence #58003 to assert the corrected behavior and adds new coverage for main-agent, named-agent, and explicit-override scenarios

Background

PR #58003 reordered resolveSubagentConfiguredModelSelection to prefer agentConfig.model over agents.defaults.subagents.model. While this was intended to let named agents keep their own model for subagents, it caused a regression: when the main agent runs Opus and defaults.subagents.model is GPT-5.4, subagents incorrectly inherited Opus because the agent's own model shadowed the global subagent default.

The correct precedence is:

agentConfig.subagents.model — explicit per-agent subagent config (highest)
agents.defaults.subagents.model — global subagent default
agentConfig.model — agent's own model (last-resort fallback)

Test plan

Real behavior proof

Behavior addressed: Subagent model resolution for a main agent whose own model differs from agents.defaults.subagents.model. Before this PR, an Opus-running agent with defaults.subagents.model = openai/gpt-5 still produced Opus subagents because the agent's own model shadowed the global subagent default in resolveSubagentConfiguredModelSelection, in the shared resolveSubagentModelConfigSelectionResult helper, and in the cron isolated-agent path. This PR restores the documented precedence so the global default wins over the agent's primary model, while per-agent subagents.model still takes the very top slot.

Real environment tested: Linux x86_64 (Node v24.15.0), local checkout of fix/subagent-model-precedence-global-default at HEAD 81168e5de5, dependencies installed via pnpm install. Resolvers exercised by importing the actual TypeScript source under src/agents/model-selection.ts and src/agents/agent-scope.ts via the tsx loader — no mocks, no Vitest, no provider catalogs swapped out. The cron isolated-agent path uses the same resolveSubagentModelConfigSelectionResult helper, so the asserted source values cover that resolver too.

Exact steps or command run after this patch:

$ cd pr-58823
$ pnpm install --prefer-offline
$ node --import tsx scripts/proof/pr-58823-subagent-precedence.mjs

The script (scripts/proof/pr-58823-subagent-precedence.mjs, committed in this PR for reviewer reproducibility) constructs three real OpenClawConfig shapes, calls resolveSubagentModelConfigSelectionResult, resolveSubagentConfiguredModelSelection, and resolveSubagentSpawnModelSelection on each, and asserts both the resolved model ref and the selection source label (subagent / default-subagent / agent).

Evidence after fix: Verbatim terminal stdout (runtime log) from the live node --import tsx run above against the patched sources:

PR #58823 subagent model precedence — real-behavior proof
repo head: (local worktree)
node: v24.15.0

--- scenario A — global default beats agent primary ---
input.agentConfig.model              = anthropic/claude-opus
input.defaults.subagents.model       = openai/gpt-5
input.agentConfig.subagents.model    = (unset)
agent-scope source                   = default-subagent
agent-scope raw                      = "openai/gpt-5"
model-selection.resolveConfigured()  = openai/gpt-5
model-selection.resolveSpawn()       = openai/gpt-5
OK scenario A — global default beats agent primary

--- scenario B — per-agent override wins ---
input.agentConfig.model              = anthropic/claude-opus
input.defaults.subagents.model       = openai/gpt-5
input.agentConfig.subagents.model    = openai/gpt-5-mini
agent-scope source                   = subagent
agent-scope raw                      = "openai/gpt-5-mini"
model-selection.resolveConfigured()  = openai/gpt-5-mini
model-selection.resolveSpawn()       = openai/gpt-5-mini
OK scenario B — per-agent override wins

--- scenario C — falls through to agent primary ---
input.agentConfig.model              = anthropic/claude-opus
input.defaults.subagents.model       = (unset)
input.agentConfig.subagents.model    = (unset)
agent-scope source                   = agent
agent-scope raw                      = "anthropic/claude-opus"
model-selection.resolveConfigured()  = anthropic/claude-opus
model-selection.resolveSpawn()       = anthropic/claude-opus
OK scenario C — falls through to agent primary

PROOF PASSED — subagent model precedence follows: per-agent > defaults.subagents > agent primary

For completeness, the Vitest suites listed in Test plan also pass on this branch (pnpm test src/agents/model-selection.test.ts → 105 passed; pnpm test src/cron/isolated-agent.model-formatting.test.ts → 32 passed; etc.), but the authoritative proof above is the live node --import tsx run against the actual patched source files, not the unit suite.

Observed result after fix:

Scenario A (Opus main agent, defaults.subagents.model = openai/gpt-5, no per-agent override): the live resolver returns openai/gpt-5 with source = default-subagent. Before this PR it returned anthropic/claude-opus with source = agent — the exact bug from fix(agents): subagent model precedence — global default shadowed by parent agent's own model #58822.
Scenario B (per-agent subagents.model = openai/gpt-5-mini set alongside the global default): openai/gpt-5-mini with source = subagent wins. Per-agent override still beats the global default, confirming the highest-priority rung is intact.
Scenario C (neither per-agent subagents.model nor defaults.subagents.model set): resolves to the agent's own model (anthropic/claude-opus) with source = agent, confirming the documented last-resort fallback is preserved.

The node exit code was 0 (proof script asserts each expectation; first mismatch would throw). The agent-scope source labels are produced by the shared resolveSubagentModelConfigSelectionResult helper used by both src/agents/model-selection.ts and src/cron/isolated-agent/model-selection.ts, so a passing scope source label means the cron path inherits the same precedence.

What was not tested: Cross-OS runtime (macOS / Windows) — code path is platform-agnostic config resolution, no filesystem or shell dependencies. Live spawn of an actual subagent process against a remote model provider — the resolver returns the same model ref regardless of whether the downstream spawn succeeds, and the spawn path is exercised by src/agents/openclaw-tools.subagents.sessions-spawn.model.test.ts (11 passing). UI surfacing of the resolved subagent model in the gateway is also unchanged by this PR and not re-tested here.

🤖 Generated with Claude Code

Co-Authored-By: Claude Opus 4.6 [email protected]

greptile-apps · 2026-04-01T07:27:03Z

Greptile Summary

This PR fixes a model-selection precedence regression introduced in PR #58003. It restores the correct priority order in both resolveSubagentConfiguredModelSelection (src/agents/model-selection.ts) and the cron isolated-agent path (src/cron/isolated-agent/model-selection.ts): per-agent subagents.model → global defaults.subagents.model → agent's own model, replacing the incorrect ordering that had the agent's own model shadowing the global subagent default.

The two-line swap in each file is minimal and correct, and both paths are now consistently aligned.
Updated tests accurately capture the regression scenario (main agent on Opus + global subagent default GPT-5.4 → subagents use GPT-5.4) and add four new cases covering the full priority matrix.
No unrelated changes; no new dependencies introduced.

Confidence Score: 5/5

Safe to merge — minimal, targeted fix with good regression-test coverage and no collateral changes.

All findings are P2 or lower. The logic change is a correct two-line swap applied consistently across both code paths, the updated and new tests directly cover the regression and the full priority matrix, and no unrelated code is touched.

No files require special attention.

_{Reviews (1): Last reviewed commit: "fix(agents): restore global subagent mod..." | Re-trigger Greptile}

steipete · 2026-04-26T00:33:55Z

Codex review: keeping this open for maintainer follow-up; there is still a little grit to resolve.

Keep this PR open. Current main still implements and tests the opposite subagent model precedence, while the docs support the reporter's expectation that agents.defaults.subagents.model is the spawned sub-agent default. The PR is also paired with open same-author issue #58822 via closing syntax, and overlapping PR #72877 is still open rather than merged.

Best possible solution:

Keep this PR open for maintainer review. If the documented contract is intended, land this PR or a narrow equivalent, ideally with explicit allowlist validation for resolved subagent defaults and sessions_spawn.model; if #58003's named-agent precedence is the intended contract, update docs/tools/subagents.md and docs/gateway/config-tools.md, then close #58822 and this PR with that product decision. Also compare with open PR #72877 before choosing the canonical fix.

What I checked:

Current helper precedence is opposite of this PR: resolveSubagentConfiguredModelSelection checks per-agent subagents.model, then the agent's own model, then agents.defaults.subagents.model. That is the ordering this PR proposes to reverse. (src/agents/model-selection.ts:245, 7120f5b25487)
Cron path still mirrors current opposite precedence: The isolated cron model-selection path checks agentConfigOverride.subagents.model, then agentConfigOverride.model, then cfg.agents.defaults.subagents.model. (src/cron/isolated-agent/model-selection.ts:71, 7120f5b25487)
Checked-in tests pin the opposite contract: The current main unit test is named prefers the agent primary model over agents.defaults.subagents.model and expects the agent primary model to win over the global subagent default. (src/agents/model-selection.test.ts:1528, 7120f5b25487)
Cron regression test also pins agent model over global subagent default: The cron formatting test expects anthropic/claude-opus-4-6 from the agent override to win over ollama/llama3.2:3b from agents.defaults.subagents.model. (src/cron/isolated-agent.model-formatting.test.ts:525, 7120f5b25487)
Docs support the PR's expected default behavior: The subagent docs say the model inherits the caller unless agents.defaults.subagents.model or per-agent agents.list[].subagents.model is set, and that explicit sessions_spawn.model still wins. Public docs: docs/tools/subagents.md. (docs/tools/subagents.md:136, 7120f5b25487)
Config docs describe global subagent model as the spawned default: The config docs describe agents.defaults.subagents.model as the default model for spawned sub-agents and say omission means sub-agents inherit the caller's model. Public docs: docs/gateway/config-tools.md. (docs/gateway/config-tools.md:373, 7120f5b25487)

Remaining risk / open question:

The intended product contract is unresolved: current code/tests favor named-agent primary models, while current docs favor a configured global subagent default overriding caller inheritance.
Merging as-is should account for allowlist behavior: defaultModel is implicitly allowed, and the PR would make agents.defaults.subagents.model the effective default in more subagent-session cases.
The normal sessions_spawn model override path still resolves and persists a model string before visible allowlist/catalog validation; the Aisle findings are concrete enough for maintainer review rather than cleanup closure.
Overlapping PR fix(agents): apply subagent spawn model selection #72877 covers this precedence plus runtime model persistence, so maintainers should choose the canonical implementation path instead of closing this PR as obsolete before another fix lands.

Codex review notes: model gpt-5.5, reasoning high; reviewed against 7120f5b25487.

…t own model PR openclaw#58003 reordered resolveSubagentConfiguredModelSelection to prefer agentConfig.model over agents.defaults.subagents.model. This caused the main agent's own model to shadow the global subagent default — e.g. when the main agent runs Opus and defaults.subagents.model is GPT-5.4, subagents incorrectly inherited Opus instead of GPT-5.4. Restore the correct precedence: 1. agentConfig.subagents.model (explicit per-agent subagent config) 2. agents.defaults.subagents.model (global subagent default) 3. agentConfig.model (agent own model, last-resort fallback) Apply the same fix to the cron isolated-agent model selection path. Co-Authored-By: Claude Opus 4.6 <[email protected]>

Add regression tests for resolveSubagentSpawnModelSelection covering the no-subagents-config 3-tier precedence: 1. session-level modelOverride wins over everything 2. agents.list[].model wins over agents.defaults.model 3. agents.defaults.model is used when neither override is set These complement the resolveSubagentConfiguredModelSelection tests added in the prior fix commit by pinning the spawn-time resolution path. Co-Authored-By: Claude Opus 4.7 (1M context) <[email protected]>

clawsweeper · 2026-04-29T18:54:26Z

Codex review: needs maintainer review before merge. Reviewed July 12, 2026, 12:13 AM ET / 04:13 UTC.

Summary
The PR makes the shared subagent configuration selector prefer the global subagent default over an agent primary, updates agent and isolated-cron regression coverage, and adds a direct resolver proof script.

PR surface: Source +7, Tests +251, Other +127. Total +385 across 7 files.

Reproducibility: yes. at source level: current main contains two deterministic selectors with opposite precedence for the same configured agent primary and global subagent default.

Review metrics: 1 noteworthy metric.

Config precedence: 1 shipped default relationship reversed. Existing configurations containing both values can silently select a different provider and model after upgrade.

Stored data model
Persistent data-model change detected: vector/embedding metadata: src/agents/agent-scope.test.ts. Confirm migration or upgrade compatibility proof before merge.

Root-cause cluster
Relationship: fixed_by_candidate
Canonical: #58822
Summary: The paired issue is canonical for the precedence mismatch; this PR is its narrow candidate fix, while the broader open PR contains additional runtime-session behavior.

Members:

canonical: fix(agents): subagent model precedence — global default shadowed by parent agent's own model #58822 - The issue reports the exact global-default-versus-agent-primary precedence mismatch fixed by this branch.
partial_overlap: fix(subagent): resolve runtime model from subagent default instead of parent primary #72984 - It includes the same precedence correction plus distinct session persistence and inherited parent-override fixes.
partial_overlap: Fix subagent default model precedence #81783 - This merged PR implemented the sessions_spawn planning half of global-default-first behavior.
partial_overlap: refactor(agents): share subagent cron fallback selection #82328 - This merged PR centralized cron fallback selection around the shared helper that still has the opposite ordering.

Proposal only: this assessment does not dispatch repair, suppress jobs, mutate sibling items, close, or merge anything.

Merge readiness
Overall: 🐚 platinum hermit
Proof: 🦞 diamond lobster
Patch quality: 🐚 platinum hermit
Result: ready for maintainer review.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

[P2] Obtain the likely owner’s explicit precedence decision and confirm this PR as the canonical narrow selector fix.

Risk before merge

[P1] Merging reverses a previously shipped routing contract for named agents and isolated cron runs when both an agent primary and global subagent default are configured.
[P2] Affected upgrades can silently switch provider credentials, billing path, fallback policy, and model-scoped runtime policy without any configuration migration or operator action.

Maintainer options:

Approve global-default precedence (recommended)
Accept the compatibility change and land this focused shared-selector alignment with its existing regression coverage.
Preserve named-agent routing
Pause or close this PR and align sessions_spawn plus documentation with the agent-primary-first contract.
Specify split behavior first
Pause overlapping work until maintainers define separate default-agent and named-agent precedence rules.

Next step before merge

[P2] The remaining action is an explicit owner decision on a shipped compatibility contract and canonical scope, not a mechanical defect suitable for automated repair.

Maintainer decision needed

Question: Should agents.defaults.subagents.model override a named agent’s primary model across all native subagent and isolated-cron paths?
Rationale: Both opposite precedence contracts have been merged into shipped code, and choosing either changes existing users’ provider/model routing; implementation review cannot safely infer the intended compatibility policy.
Likely owner: steipete — He authored the central shared cron/fallback refactor and has already reviewed the exact contract conflict across the affected paths.
Options:
- Global default wins everywhere (recommended): Merge this narrow shared-selector alignment while retaining per-agent subagents.model as the only higher-priority configured override.
- Named agent primary wins: Reject this PR and revise the spawn planner and documentation to restore agent-primary-first behavior consistently.
- Split main and named semantics: Pause both overlapping PRs and define distinct precedence for the default agent and named agents before implementation.

Security
Cleared: The patch adds no dependency, workflow permission, secret handling, artifact download, publishing, or new executable trust boundary.

Review details

Best possible solution:

Adopt global-default-first as one documented contract across spawn planning, shared cron selection, and fallback policy, then assess the broader runtime-session PR only for its distinct persistence and parent-override fixes.

Do we have a high-confidence way to reproduce the issue?

Yes at source level: current main contains two deterministic selectors with opposite precedence for the same configured agent primary and global subagent default.

Is this the best way to solve the issue?

Yes mechanically: reordering the shared helper is the narrowest maintainable fix for the remaining inconsistency, provided maintainers approve global-default-first as the compatibility contract.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against 86b10ffb8e93.

Label changes

Label justifications:

P2: This is a bounded but user-visible model-routing and cost compatibility issue affecting subagent and isolated-cron workflows.
merge-risk: 🚨 compatibility: The patch reverses behavior that was reachable from shipped releases for existing named-agent configurations.
merge-risk: 🚨 auth-provider: The selected provider, credentials, billing route, fallbacks, and model-scoped runtime policy can change after merge.
rating: 🐚 platinum hermit: Overall readiness is 🐚 platinum hermit; proof is 🦞 diamond lobster and patch quality is 🐚 platinum hermit.
status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Sufficient (terminal): The PR provides after-fix terminal output from the actual selector functions covering the global default, per-agent override, and agent fallback; the exact head retains the same selector behavior and CI includes a successful real-behavior-proof check.
proof: sufficient: Contributor real behavior proof is sufficient. The PR provides after-fix terminal output from the actual selector functions covering the global default, per-agent override, and agent fallback; the exact head retains the same selector behavior and CI includes a successful real-behavior-proof check.

Evidence reviewed

PR surface:

Source +7, Tests +251, Other +127. Total +385 across 7 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	2	9	2	+7
Tests	4	264	13	+251
Docs	0	0	0	0
Config	0	0	0	0
Generated	0	0	0	0
Other	1	127	0	+127
Total	7	400	15	+385

What I checked:

Current-main selector mismatch: The sessions_spawn selector uses per-agent subagent model, then global subagent default, then agent primary, while the shared cron/fallback selector still places agent primary before the global default. (src/agents/model-selection-config.ts:44, 86b10ffb8e93)
Meaningful exact-head production change: Against current main, the branch changes the shared selector used by isolated cron model and fallback resolution; the remaining production behavior is not already implemented on main. (src/agents/agent-scope.ts:461, 913d0ba36379)
Documented behavior: Current user documentation describes agents.defaults.subagents.model as the spawned subagent default and says caller inheritance applies only when it is omitted. Public docs: docs/tools/subagents.md. (docs/tools/subagents.md:143, 86b10ffb8e93)
Earlier named-agent contract provenance: Merged commit e394262 intentionally reordered both native and isolated-cron selection so a named agent primary outranked the global subagent default. (src/agents/model-selection.ts:440, e394262bd87e)
Later global-default contract provenance: Merged commit 2b04ced reversed the sessions_spawn selector so the global subagent default wins and its model-scoped runtime policy remains attached. (src/agents/model-selection-config.ts:44, 2b04cedfb131)
Shared cron ownership provenance: Merged commit 0ad3d25 centralized isolated-cron primary and fallback policy on the shared selector while retaining agent-primary-first ordering there. (src/cron/isolated-agent/model-selection.ts:83, 0ad3d25fb7cd)

Likely related people:

steipete: Authored the merged cron/fallback refactor around the selector at issue and previously documented this exact cross-surface compatibility choice in the PR discussion. (role: introduced shared cron selection and recent reviewer; confidence: high; commits: 0ad3d25fb7cd; files: src/agents/agent-scope.ts, src/cron/isolated-agent/model-selection.ts, src/cron/isolated-agent/run-fallback-policy.ts)
joshavant: Authored the merged change that made the global subagent default win in sessions_spawn planning and preserved model-scoped runtime policy. (role: introduced current spawn-planner behavior; confidence: high; commits: 2b04cedfb131; files: src/agents/model-selection-config.ts, src/agents/model-selection.test.ts)
neeravmakwana: Authored the merged March change that intentionally made the agent primary outrank the global subagent default in native and cron selection. (role: introduced earlier named-agent behavior; confidence: high; commits: e394262bd87e; files: src/agents/model-selection.ts, src/cron/isolated-agent/model-selection.ts)

What the crustacean ranks mean

🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works

ClawSweeper keeps one durable marker-backed review comment per issue or PR.
Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
Maintainers can also comment @clawsweeper review to request a fresh review only.
Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

Review history (6 earlier review cycles)

reviewed 2026-07-03T05:58:41.033Z sha 02ec76d :: needs maintainer review before merge. :: none
reviewed 2026-07-05T16:30:08.166Z sha 02ec76d :: needs maintainer review before merge. :: none
reviewed 2026-07-11T19:47:38.949Z sha a258165 :: needs maintainer review before merge. :: none
reviewed 2026-07-11T20:17:45.527Z sha a258165 :: needs maintainer review before merge. :: none
reviewed 2026-07-11T20:26:37.900Z sha a258165 :: needs maintainer review before merge. :: none
reviewed 2026-07-11T20:47:54.419Z sha 49b4d13 :: needs maintainer review before merge. :: none

…precedence-global-default

…ack tests PR openclaw#58823's subagent-precedence fallback tests (agent-scope + cron isolated-agent) asserted legacy 'openai-codex/gpt-5.x', but their configs use 'openai/gpt-5.x' and the resolver returns the configured fallbacks verbatim (no aliasing) — so the expected value cannot be openai-codex. Per AGENTS.md, openai-codex is folded into openai (legacy input only), so the canonical runtime value is openai/*. The flip behavior (global defaults.subagents.model fallbacks win over the agent's own model) is correctly implemented in agent-scope.ts; only the test expectations were wrong. Fixes the 2 failing CI shards (agentic-agents-core-runtime, core-runtime-cron-isolated-agent).

…ecedence-global-default

joeykrug · 2026-07-11T20:08:37Z

Refreshed the canonical branch with upstream main at af1505df659f80893829be5c96e6e9918a9fb42d. Exact head a2581657124f8f6ca7cd1d861bcbe0280af1a49f is green across CI, CodeQL, CodeQL Critical Quality, OpenGrep, Workflow Sanity, and iOS Periphery. The remaining precedence choice is the maintainer product decision already documented. @clawsweeper re-review

joeykrug · 2026-07-11T20:11:31Z

@clawsweeper re-review

clawsweeper · 2026-07-11T20:11:33Z

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

Re-review progress:

State: Complete
Detail: The targeted re-review published an exact durable review comment and queued an executing target-wide serialized router scan.
Run: https://github.com/openclaw/clawsweeper/actions/runs/29166816231
Updated: 2026-07-11T20:28:14.766Z

joeykrug · 2026-07-11T20:34:30Z

Final shared-upstream synchronization is exact head 49b4d13321fcda6e3ca3ef2744f17a0dca52749a, a semantic true merge containing cutoff 262baedcf7fa3d66dc35cce7d0253e0a31f0cc0e while preserving main's current agent-scope changes and the documented precedence correction.

@clawsweeper re-review

clawsweeper · 2026-07-11T20:34:33Z

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

Re-review progress:

State: Complete
Detail: The targeted re-review published an exact durable review comment and queued an executing target-wide serialized router scan.
Run: https://github.com/openclaw/clawsweeper/actions/runs/29167459844
Updated: 2026-07-11T20:49:36.173Z

openclaw-barnacle Bot added agents Agent runtime and tooling size: S labels Apr 1, 2026

joeykrug force-pushed the fix/subagent-model-precedence-global-default branch 3 times, most recently from 155066c to d588fb4 Compare April 8, 2026 04:50

joeykrug force-pushed the fix/subagent-model-precedence-global-default branch from 9a24cf6 to 0cc6234 Compare April 26, 2026 04:25

joeykrug and others added 2 commits April 26, 2026 13:42

joeykrug force-pushed the fix/subagent-model-precedence-global-default branch from 4645342 to 5f5e960 Compare April 26, 2026 18:13

openclaw-barnacle Bot added size: M and removed size: S labels Apr 26, 2026

Merge branch 'main' into fix/subagent-model-precedence-global-default

55f4bf3

Papilllion mentioned this pull request Apr 27, 2026

fix(agents): apply subagent spawn model selection #72877

Closed

joeykrug added 3 commits April 27, 2026 17:47

Merge branch 'main' into fix/subagent-model-precedence-global-default

433be57

Merge branch 'main' into fix/subagent-model-precedence-global-default

ea1811e

Merge branch 'main' into fix/subagent-model-precedence-global-default

e74c7ab

clawsweeper Bot mentioned this pull request Apr 28, 2026

fix(agents): subagent model precedence — global default shadowed by parent agent's own model #58822

Open

Merge branch 'main' into fix/subagent-model-precedence-global-default

8b2e7f5

Merge branch 'main' into fix/subagent-model-precedence-global-default

271777b

This was referenced May 1, 2026

fix(subagent): resolve runtime model from subagent default instead of parent primary #72984

Open

docs: subagent model priority order mismatch between docs and implementation #75934

Closed

joeykrug added 2 commits May 4, 2026 14:08

Merge branch 'main' into fix/subagent-model-precedence-global-default

ec17d32

Merge branch 'main' into fix/subagent-model-precedence-global-default

581b6e3

openclaw-barnacle Bot added the triage: needs-real-behavior-proof Candidate: external PR needs after-fix proof from a real setup. label May 5, 2026

clawsweeper Bot mentioned this pull request May 9, 2026

Subagent model default ignored: parent session model overrides agents.defaults.subagents.model #79888

Closed

joeykrug added 2 commits May 9, 2026 18:11

Merge branch 'main' into fix/subagent-model-precedence-global-default

56e9a9f

Merge branch 'main' into fix/subagent-model-precedence-global-default

14735c1

Merge remote-tracking branch 'upstream/main' into fix/subagent-model-…

7420806

…precedence-global-default

openclaw-barnacle Bot removed the proof: sufficient ClawSweeper judged the real behavior proof convincing. label Jun 3, 2026

clawsweeper Bot added the proof: sufficient ClawSweeper judged the real behavior proof convincing. label Jun 3, 2026

Merge branch 'main' into fix/subagent-model-precedence-global-default

ac59fd4

openclaw-barnacle Bot removed the proof: sufficient ClawSweeper judged the real behavior proof convincing. label Jun 4, 2026

Merge branch 'main' into fix/subagent-model-precedence-global-default

63b0013

openclaw-barnacle Bot removed extensions: memory-wiki extensions: deepinfra labels Jun 8, 2026

joeykrug added 2 commits June 11, 2026 12:03

Merge remote-tracking branch 'origin/main' into fix/subagent-model-pr…

d5ec8f1

…ecedence-global-default

Merge branch 'main' into fix/subagent-model-precedence-global-default

dc03657

This was referenced Jun 19, 2026

[Bug]: Sub-agent model routing ignores model parameter, silently falls back to deepseek #91171

Closed

fix(agents): persist modelOverride/providerOverride in subagent spawn session patch #95210

Closed

joeykrug added 3 commits June 28, 2026 18:30

Merge remote-tracking branch 'origin/main' into maint/pr-58823

d95a384

Merge branch 'main' into fix/subagent-model-precedence-global-default

02ec76d

Merge upstream/main into fix/subagent-model-precedence-global-default

a258165

Merge final upstream/main cutoff into subagent precedence fix

49b4d13

Merge branch 'main' into fix/subagent-model-precedence-global-default

913d0ba

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

fix(agents): restore global subagent model default priority over agent own model#58823

fix(agents): restore global subagent model default priority over agent own model#58823
joeykrug wants to merge 35 commits into
openclaw:mainfrom
joeykrug:fix/subagent-model-precedence-global-default

joeykrug commented Apr 1, 2026 •

edited

Loading

Uh oh!

greptile-apps Bot commented Apr 1, 2026

Uh oh!

steipete commented Apr 26, 2026 •

edited by clawsweeper Bot

Loading

Uh oh!

clawsweeper Bot commented Apr 29, 2026 •

edited

Loading

Uh oh!

joeykrug commented Jul 11, 2026

Uh oh!

joeykrug commented Jul 11, 2026

Uh oh!

clawsweeper Bot commented Jul 11, 2026 •

edited

Loading

Uh oh!

joeykrug commented Jul 11, 2026

Uh oh!

clawsweeper Bot commented Jul 11, 2026 •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

Uh oh!

Uh oh!

Conversation

joeykrug commented Apr 1, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Current exact-head refresh

Background

Test plan

Real behavior proof

Uh oh!

greptile-apps Bot commented Apr 1, 2026

Greptile Summary

Confidence Score: 5/5

Uh oh!

steipete commented Apr 26, 2026 • edited by clawsweeper Bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented Apr 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

joeykrug commented Jul 11, 2026

Uh oh!

joeykrug commented Jul 11, 2026

Uh oh!

clawsweeper Bot commented Jul 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

joeykrug commented Jul 11, 2026

Uh oh!

clawsweeper Bot commented Jul 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

joeykrug commented Apr 1, 2026 •

edited

Loading

steipete commented Apr 26, 2026 •

edited by clawsweeper Bot

Loading

clawsweeper Bot commented Apr 29, 2026 •

edited

Loading

clawsweeper Bot commented Jul 11, 2026 •

edited

Loading

clawsweeper Bot commented Jul 11, 2026 •

edited

Loading