fix: honor exec approval security and clean up stale tasks by lml2468 · Pull Request #60310 · openclaw/openclaw

lml2468 · 2026-04-03T13:23:33Z

Summary

This PR now contains two verified fixes:

Gateway exec now honors exec-approvals.json agent security directly, matching the node-host path.
Task maintenance now correctly marks stale cron runs and CLI tasks backed only by long-lived chat sessions as lost.

The earlier cron wake fallback experiment was removed from this PR.

Changes

Exec approvals (`#60268`)

src/agents/bash-tools.exec-host-shared.ts

change gateway hostSecurity to use approvals.agent.security directly instead of intersecting it with the tool default via minSecurity(...)
keep existing ask precedence logic unchanged
add a focused regression test for the broader-security case

Why this is correct:

exec-approvals.json is the explicit user security policy
node-host already treated the resolved approval security as authoritative
when no agent/default override exists, approvals.agent.security falls back to params.security, so behavior stays backward-compatible

Task maintenance (`#60299`)

src/tasks/task-registry.maintenance.ts

treat runtime="cron" tasks as having no persistent backing session after the run exits
treat runtime="cli" tasks with channel/group/direct child session keys as stale even if that long-lived chat session still exists in the session store
keep subagent and acp behavior intact

src/tasks/task-registry.maintenance.issue-60299.test.ts

add regression coverage for stale cron tasks, stale CLI tasks backed by persistent chat sessions, and the healthy subagent case

Verification

pnpm check
pnpm test src/tasks/task-registry.maintenance.issue-60299.test.ts

I also added a targeted exec regression test in-tree. A standalone local runner for that file was sticky in this environment, so the landed proof for the exec path is the code change plus that focused unit coverage.

greptile-apps · 2026-04-03T13:26:15Z

Greptile Summary

This PR bundles three bug fixes across different subsystems: (1) the main fix aligns gateway exec security with node-host by making exec-approvals.json authoritative for hostSecurity instead of intersecting it with the tool-default via minSecurity; (2) a cron timer fix ensures requestHeartbeatNow is called when runHeartbeatOnce is skipped for a non-busy reason, preventing enqueued system events from silently going unconsumed; (3) hasBackingSession in task maintenance is fixed to correctly handle cron tasks (always stale) and cli tasks whose childSessionKey points to a persistent channel/group/direct session.

The security fix is backward-compatible: when no explicit entry exists in exec-approvals.json, approvals.agent.security resolves through resolveExecApprovalsFromFile back to params.security (the tool/runtime default), matching pre-fix behavior. The maintenance fixes are well-covered by the new regression test file.

Confidence Score: 5/5

Safe to merge — all three fixes are backward-compatible, well-commented, and the maintenance path has direct regression test coverage.

No P0 or P1 findings. The exec security fix is provably backward-compatible (verified via resolveExecApprovalsFromFile fallback chain). The timer change follows the same requestHeartbeatNow pattern already used elsewhere in the same function. The hasBackingSession refactor correctly handles all four TaskRuntime values. Tests cover the three newly fixed stale-task cases plus a healthy subagent non-regression case.

No files require special attention.

_{Reviews (1): Last reviewed commit: "fix: subagents now inherit exec approval..." | Re-trigger Greptile}

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 73e9a2b1c9

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

chatgpt-codex-connector · 2026-04-03T13:28:28Z

+  if (task.runtime === "cron") {
+    return false;
+  }


Stop treating every active cron task as orphaned

Returning false for all runtime === "cron" tasks makes shouldMarkLost() reconcile any running cron task to lost once the 5-minute grace expires, even when the job is still executing normally. This is a real runtime path because cron agent turns are allowed to run far longer (default safety timeout is up to 60 minutes), so long jobs will be mislabeled as lost mid-run and can emit incorrect terminal state/notifications before they actually finish.

Useful? React with 👍 / 👎.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: c4b1bcdff7

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-04-04T06:40:45Z

+    const chatType = deriveSessionChatType(childSessionKey);
+    if (chatType === "channel" || chatType === "group" || chatType === "direct") {
+      return false;
+    }


Preserve liveness checks for active CLI chat runs

Returning false for CLI tasks whose childSessionKey parses as channel/group/direct makes shouldMarkLost() flip them to lost after the 5-minute grace even when the run is still healthy. This path is reachable for gateway-ingress tasks (runtime: "cli") that use chat-scoped session keys, and agent runs are allowed to exceed 5 minutes (default timeout is far longer), so long-running turns will be mislabeled as terminal mid-execution and can be swept/inspected as dead work before they actually finish.

Useful? React with 👍 / 👎.

chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 258e5e7d33

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

chatgpt-codex-connector · 2026-04-04T07:45:44Z

+function resolveBaselineFloor(
+  provider: ProviderKey,
+  lane: BaselineLane,
+): LiveCacheFloor | undefined {
+  return (LIVE_CACHE_REGRESSION_BASELINE[provider] as Record<string, LiveCacheFloor>)[lane];


Remove duplicate resolveBaselineFloor implementation

This introduces a second resolveBaselineFloor implementation while the original one earlier in this file is still present. TypeScript treats this as duplicate function implementations, which can break pnpm tsgo and prevent importing this module (for example through scripts/check-live-cache.ts and the live cache regression test) before any runtime logic executes. Keep a single helper and change its parameter type in place instead of redefining it.

Useful? React with 👍 / 👎.

steipete · 2026-04-04T07:52:12Z

Landed via temp rebase onto main.

Gate: pnpm check, pnpm build, pnpm test src/tasks/task-registry.maintenance.issue-60299.test.ts, pnpm test src/channels/plugins/config-schema.test.ts, plus CI
Land commit: 4e424d9
Merge commit: 19036ef

Thanks @lml2468!

@lml2468

…anks @lml2468)

@lml2468

…nks @lml2468)

@lml2468

…anks @lml2468)

@lml2468

…nks @lml2468)

…lost marker Upstream 7d1575b (openclaw#60310, 2026-04-04) introduced activeJobIds plus markCronJobActive/clearCronJobActive so task-registry maintenance has a backing-session signal for runtime='cron' tasks (task-registry.maintenance.ts hasBackingSession). That patch wired the pair into runDueJob and executeJob but left the manual-run path (prepareManualRun + finishPreparedManualRun in src/cron/service/ops.ts) without mark/clear. Symptom: when a user triggers `openclaw cron run <id> --force` (or any manual-run RPC) and the run exceeds TASK_RECONCILE_GRACE_MS (5 min) — common for force-mode `agentTurn` runs which can reach AGENT_TURN_SAFETY_TIMEOUT_MS (60 min) — task-registry sweeper marks the active task `lost` and emits a `Background task lost` system message to the session, even though the run is still progressing normally. The merged commit 1fae716 (resolveDurableCronTaskRecovery, 2026-04-26) reconciles terminal status retroactively from cron run-log + store.lastRunStatus, but only after the run finishes. This patch suppresses the transient `lost` marker by adding the producer-side mark/clear pair, restoring symmetry with runDueJob/executeJob: * prepareManualRun: markCronJobActive(job.id) after tryCreateManualTaskRun. * finishPreparedManualRun: wrap body in try/finally with clearCronJobActive(jobId). Scope intentionally narrower than openclaw#71040 (closed): * No change to runStartupCatchupCandidate — the `deferAgentTurnJobs:true` policy added in 7877182 reroutes long-running startup catchups to runDueJob/executeJob (already wired). Non-agentTurn startup catchups are theoretical hot-path per pre-PR cross-review. * No change to active-jobs.ts API. Regression test (src/cron/active-jobs-manual-run.test.ts): two cases exercising production hot-path via cron.run("<id>", "force") — success and inner-throw — assert isCronJobActive transitions true→false around the manual run. Fixes openclaw#78233. [AI-assisted, fully tested]

@lml2468

…anks @lml2468)

@lml2468

…nks @lml2468)

…lost marker Upstream 7d1575b (openclaw#60310, 2026-04-04) introduced activeJobIds plus markCronJobActive/clearCronJobActive so task-registry maintenance has a backing-session signal for runtime='cron' tasks (task-registry.maintenance.ts hasBackingSession). That patch wired the pair into runDueJob and executeJob but left the manual-run path (prepareManualRun + finishPreparedManualRun in src/cron/service/ops.ts) without mark/clear. Symptom: when a user triggers `openclaw cron run <id> --force` (or any manual-run RPC) and the run exceeds TASK_RECONCILE_GRACE_MS (5 min) — common for force-mode `agentTurn` runs which can reach AGENT_TURN_SAFETY_TIMEOUT_MS (60 min) — task-registry sweeper marks the active task `lost` and emits a `Background task lost` system message to the session, even though the run is still progressing normally. The merged commit 1fae716 (resolveDurableCronTaskRecovery, 2026-04-26) reconciles terminal status retroactively from cron run-log + store.lastRunStatus, but only after the run finishes. This patch suppresses the transient `lost` marker by adding the producer-side mark/clear pair, restoring symmetry with runDueJob/executeJob: * prepareManualRun: markCronJobActive(job.id) after tryCreateManualTaskRun. * finishPreparedManualRun: wrap body in try/finally with clearCronJobActive(jobId). Scope intentionally narrower than openclaw#71040 (closed): * No change to runStartupCatchupCandidate — the `deferAgentTurnJobs:true` policy added in 7877182 reroutes long-running startup catchups to runDueJob/executeJob (already wired). Non-agentTurn startup catchups are theoretical hot-path per pre-PR cross-review. * No change to active-jobs.ts API. Regression test (src/cron/active-jobs-manual-run.test.ts): two cases exercising production hot-path via cron.run("<id>", "force") — success and inner-throw — assert isCronJobActive transitions true→false around the manual run. Fixes openclaw#78233. [AI-assisted, fully tested]

@lml2468

…anks @lml2468)

@lml2468

…nks @lml2468)

openclaw-barnacle Bot added agents Agent runtime and tooling size: M labels Apr 3, 2026

chatgpt-codex-connector Bot reviewed Apr 3, 2026

View reviewed changes

steipete changed the title ~~fix: subagents inherit exec approval security context from parent (gateway path parity with node-host)~~ fix: honor exec approval security and clean up stale tasks Apr 4, 2026

steipete force-pushed the fix/issue-60268 branch from 73e9a2b to c4b1bcd Compare April 4, 2026 06:35

chatgpt-codex-connector Bot reviewed Apr 4, 2026

View reviewed changes

steipete force-pushed the fix/issue-60268 branch from c4b1bcd to b19e07c Compare April 4, 2026 07:04

openclaw-barnacle Bot added the extensions: anthropic label Apr 4, 2026

steipete force-pushed the fix/issue-60268 branch from 88e11a2 to 0126ab9 Compare April 4, 2026 07:38

steipete added 5 commits April 4, 2026 08:39

fix: honor exec approval security from approvals (openclaw#60310)

37a213a

fix: clean up stale cron and chat-backed tasks (openclaw#60310)

0e76b59

fix: restore channel sdk schema typing

1c097ee

chore: retrigger ci

599cf1c

fix: restore main type surfaces

258e5e7

steipete force-pushed the fix/issue-60268 branch from 0126ab9 to 258e5e7 Compare April 4, 2026 07:40

openclaw-barnacle Bot removed the extensions: anthropic label Apr 4, 2026

chatgpt-codex-connector Bot reviewed Apr 4, 2026

View reviewed changes

fix: unblock current main checks

4e424d9

openclaw-barnacle Bot added the commands Command implementations label Apr 4, 2026

steipete merged commit 19036ef into openclaw:main Apr 4, 2026
9 checks passed

steipete added a commit that referenced this pull request Apr 4, 2026

fix: honor exec approval security from approvals (#60310)

6afdf10

steipete added a commit that referenced this pull request Apr 4, 2026

fix: clean up stale cron and chat-backed tasks (#60310)

f6df3ed

lovewanwan pushed a commit to lovewanwan/openclaw that referenced this pull request Apr 28, 2026

fix: honor exec approval security from approvals (openclaw#60310)

1e0f0b7

lovewanwan pushed a commit to lovewanwan/openclaw that referenced this pull request Apr 28, 2026

fix: clean up stale cron and chat-backed tasks (openclaw#60310)

ed5e2a2

lovewanwan pushed a commit to lovewanwan/openclaw that referenced this pull request Apr 28, 2026

fix: honor exec approval security from approvals (openclaw#60310) (th…

3bbfb36

…anks @lml2468)

lovewanwan pushed a commit to lovewanwan/openclaw that referenced this pull request Apr 28, 2026

fix: reconcile stale cron and chat-backed tasks (openclaw#60310) (tha…

0b3ee8e

…nks @lml2468)

ogt-redknie pushed a commit to ogt-redknie/OPENX that referenced this pull request May 2, 2026

fix: honor exec approval security from approvals (openclaw#60310)

f8b5944

ogt-redknie pushed a commit to ogt-redknie/OPENX that referenced this pull request May 2, 2026

fix: clean up stale cron and chat-backed tasks (openclaw#60310)

df06875

ogt-redknie pushed a commit to ogt-redknie/OPENX that referenced this pull request May 2, 2026

fix: honor exec approval security from approvals (openclaw#60310) (th…

3ea158e

…anks @lml2468)

ogt-redknie pushed a commit to ogt-redknie/OPENX that referenced this pull request May 2, 2026

fix: reconcile stale cron and chat-backed tasks (openclaw#60310) (tha…

efc4c5a

…nks @lml2468)

clawsweeper Bot mentioned this pull request May 2, 2026

False-positive lost cron task records after gateway restart due to transient activeJobIds backing-session check #68191

Closed

This was referenced May 6, 2026

cron: transient 'lost' marker on long-running manual runs before sweeper recovery #78233

Closed

fix(cron): mark active-jobs on manual-run path to suppress transient lost marker #78243

Merged

steipete mentioned this pull request May 7, 2026

[Bug] exec-approval-followup tasks stuck in 'running' forever, blocking channel reload and saturating event loop #76162

Closed

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request May 9, 2026

fix: honor exec approval security from approvals (openclaw#60310)

e53fbd4

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request May 9, 2026

fix: clean up stale cron and chat-backed tasks (openclaw#60310)

0b5af7a

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request May 9, 2026

fix: honor exec approval security from approvals (openclaw#60310) (th…

2eb2f9b

…anks @lml2468)

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request May 9, 2026

fix: reconcile stale cron and chat-backed tasks (openclaw#60310) (tha…

0d60f17

…nks @lml2468)

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request May 24, 2026

fix: honor exec approval security from approvals (openclaw#60310)

d353c72

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request May 24, 2026

fix: clean up stale cron and chat-backed tasks (openclaw#60310)

1248825

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request May 24, 2026

fix: honor exec approval security from approvals (openclaw#60310) (th…

7e14133

…anks @lml2468)

github-actions Bot pushed a commit to Desicool/openclaw that referenced this pull request May 24, 2026

fix: reconcile stale cron and chat-backed tasks (openclaw#60310) (tha…

a2a0636

…nks @lml2468)

clawsweeper Bot mentioned this pull request May 25, 2026

fix(tasks): recover terminal lost cron rows #86088

Open

yetval mentioned this pull request Jun 9, 2026

Bug: startup catch-up cron runs never set the active marker, so long command jobs are reconciled as lost while still running #91695

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

fix: honor exec approval security and clean up stale tasks#60310

fix: honor exec approval security and clean up stale tasks#60310
steipete merged 6 commits into
openclaw:mainfrom
lml2468:fix/issue-60268

lml2468 commented Apr 3, 2026 •

edited by steipete

Loading

Uh oh!

greptile-apps Bot commented Apr 3, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Apr 3, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Apr 4, 2026

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Apr 4, 2026

Uh oh!

Uh oh!

steipete commented Apr 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

lml2468 commented Apr 3, 2026 • edited by steipete Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Changes

Exec approvals (#60268)

Task maintenance (#60299)

Verification

Uh oh!

greptile-apps Bot commented Apr 3, 2026

Greptile Summary

Confidence Score: 5/5

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Apr 3, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Apr 4, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Apr 4, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

steipete commented Apr 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

lml2468 commented Apr 3, 2026 •

edited by steipete

Loading

Exec approvals (`#60268`)

Task maintenance (`#60299`)