Skip to content

[Bug]: Slack channel silently drops after upgrade to 2026.6.5 — plugins.allow regression #91873

Closed

Description

@nickshippers84

Bug type

Regression (worked before, now fails)

Beta release blocker

No

Summary

Version: 2026.4.26 → 2026.6.5Platform: Linux (Bluehost VPS), Slack channel via Socket Mode

What happened:
After updating to 2026.6.5 and restarting the gateway, the Slack channel silently stopped loading. Gateway reported "ready" with no errors. No Slack socket connection was established. The channel had been working fine on 2026.4.26.

Root cause (found manually):
The Slack plugin is installed at ~/.openclaw/npm/projects/openclaw-slack-*/. On 2026.4.26 it auto-loaded on startup. On 2026.6.5 it no longer auto-discovers from that path — only plugins in the extensions/ directory are picked up. Fix was to manually add plugins.allow: ["slack"] to config.

Two asks:

  1. Either restore backward-compat auto-discovery for npm-installed plugins, or document this as a breaking change with a migration step.
  2. If a configured channel (one with tokens/credentials in config) fails to load on startup, surface a loud warning — not a silent omission. Users have no way to know their channel is down until they try to use it.

Steps to reproduce

Repro path:

  1. Slack channel configured and working on 2026.4.26; plugins.allow not set in config
  2. npm install -g openclaw@latest → upgrades to 2026.6.5
  3. openclaw gateway restart
  4. Observe: startup log shows http server listening (N plugins: ...) — slack absent from list; no slack socket mode connected log entry; channel status shows SETUP

Confirmed fix: openclaw config set plugins.allow '["slack"]' + restart → slack appears in loaded plugins, socket connects.

Expected behavior

On 2026.4.26, Slack auto-loaded and connected without plugins.allow set. The expected behavior is that upgrading to 2026.6.5 preserves that — Slack loads on startup without requiring manual config changes.

Actual behavior

After upgrading to 2026.6.5 and restarting, Slack silently failed to load. Gateway logged http server listening (10 plugins: ...) with slack absent from the list. No slack socket mode connected entry appeared. Channel status showed SETUP. User received no messages and had no indication the channel was down until attempting to send.

OpenClaw version

2026.4.26 → 2026.6.5

Operating system

Linux — CentOS Stream 9 (5.14.0-687.12.1.el9_8.x86_64)

Install method

npm install -g openclaw — global npm install, running as a systemd user service on a Bluehost VPS.

Model

anthropic/claude-sonnet-4-6

Provider / routing chain

Direct — Anthropic API (api.anthropic.com), no proxy or router.

Additional provider/model setup details

No response

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response

Activity

  1. clawsweeper commented on Jun 10, 2026

    @clawsweeper
    Contributor

    Codex review: keeping this open for maintainer follow-up; there is still a little grit to resolve. Reviewed June 21, 2026, 3:36 PM ET / 19:36 UTC.

    Summary
    Keep open: current main and the latest release still require explicit trust for configured non-bundled/global channel plugins, and the only direct fix attempts for this Slack upgrade regression are closed unmerged PRs.

    Reproducibility: yes. Source inspection shows a configured global/non-bundled channel plugin without explicit trust is blocked from startup, and the startup logger does not surface that blocker; I did not run a live Slack upgrade.

    Root-cause cluster
    Relationship: canonical
    Canonical: #91873
    Summary: This issue remains the canonical tracker for the Slack/global-plugin explicit-trust upgrade regression; related direct fix attempts are closed unmerged, and generic open plugin-warning work does not own the configured Slack startup blocker.

    Members:

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

    Next step
    Manual review is needed before another fix PR because maintainers must choose between restoring implicit external-plugin loading and keeping fail-closed trust with migration/startup diagnostics.

    Review details

    Best possible solution:

    Keep the fail-closed plugin trust boundary only if upgrade, doctor, startup, or status paths make blocked configured external channels impossible to miss; otherwise add a bounded compatibility migration or activation path for upgraded configured channels.

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

    Yes. Source inspection shows a configured global/non-bundled channel plugin without explicit trust is blocked from startup, and the startup logger does not surface that blocker; I did not run a live Slack upgrade.

    Is this the best way to solve the issue?

    Unclear on exact fix shape. Automatic re-enablement would address the compatibility regression but needs security/product approval; a migration plus startup/status warning is the safer path if fail-closed trust remains intentional.

    AGENTS.md: found and applied where relevant.

    Remaining risk / open question:

    • Restoring implicit loading for globally installed external plugins could weaken the explicit plugin trust boundary unless the compatibility path is narrow and intentional.
    • Keeping the fail-closed trust policy without a migration, startup warning, or status warning leaves configured channel outages easy to miss after upgrade.
    • No live Slack Socket Mode upgrade was run in this read-only review; confidence comes from source, tests, release inspection, and related PR state.

    Codex review notes: model internal, reasoning high; reviewed against 6daf9307e0b9.

    Label changes

    Label changes:

    • remove clawsweeper:fix-shape-clear: Current issue advisory state no longer selects this label.

    Label justifications:

    • P1: A recent upgrade can leave a configured Slack messaging workflow unloaded without an operator-visible startup failure.
    • impact:message-loss: When Slack does not load, inbound Slack messages and replies can be missed until the operator notices the channel is down.
    • impact:other: The issue also affects external plugin startup and upgrade compatibility, a maintainer-visible surface outside the narrower owned impact labels.
    Evidence reviewed

    What I checked:

    • Current startup gate still requires explicit trust: For non-bundled configured channel plugins, startup planning returns true only when the activation state is enabled and explicitly enabled; a global Slack plugin is not loaded merely because channels.slack is configured. (src/plugins/gateway-startup-plugin-ids.ts:1861, 6daf9307e0b9)
    • Current configured-channel policy marks untrusted global/config owners blocked: evaluateEffectiveChannelPlugin returns blockedReason: "untrusted-plugin" for global or config-origin non-bundled channel owners unless the manifest owner is explicitly trusted. (src/plugins/channel-presence-policy.ts:251, 6daf9307e0b9)
    • Gateway startup log still lacks configured-channel blocker warnings: logGatewayStartup emits the model line, ready banner, log-file line, Nix notice, and dangerous-config warnings, but it does not call the configured-channel blocker scanner. (src/gateway/server-startup-log.ts:59, 6daf9307e0b9)
    • Doctor diagnostics know this blocker but are not startup behavior: Doctor can format external plugin "<id>" is installed without explicit trust warnings for configured channel blockers, which supports the source reproduction but does not satisfy the gateway-startup warning ask. (src/commands/doctor/shared/channel-plugin-blockers.ts:337, 6daf9307e0b9)
    • Adjacent tests source-reproduce the gate: The startup-planning tests assert that untrusted global activation owners are filtered out, while explicitly enabled global owners are eligible; doctor tests assert external configured channel plugins are reported as missing explicit enablement. (src/plugins/channel-plugin-ids.test.ts:2915, 6daf9307e0b9)
    • Latest release still has the same behavior: The v2026.6.9 source still has the same explicit-trust startup gate and no configured-channel blocker warning in the startup logger, so this is not fixed in the latest release. (src/plugins/gateway-startup-plugin-ids.ts:1861, c645ec4555c0)

    Likely related people:

    • vincentkoc: Authored and merged the manifest-driven channel startup planning and activation provenance work that currently owns the trust gate involved in this issue. (role: recent activation and startup policy contributor; confidence: high; commits: b7b384679379, f911bbc35329, d2ce3e9accc6; files: src/plugins/gateway-startup-plugin-ids.ts, src/plugins/channel-presence-policy.ts, src/plugins/channel-plugin-ids.test.ts)
    • obviyus: Authored merged PR fix: preserve bundled channel plugin compat #58873, which added doctor warnings for real channel-plugin blockers and preserved bundled channel allowlist compatibility. (role: adjacent diagnostics and bundled-channel compatibility contributor; confidence: medium; commits: fb28b0254062; files: src/commands/doctor/shared/channel-plugin-blockers.ts, src/commands/doctor/shared/channel-plugin-blockers.test.ts)
    • Peter Steinberger: Recent history on the gateway startup plugin-loading surface includes tightening startup plugin loading and preserving allowlist guards for auto-enabled bundled channels. (role: adjacent startup-loading contributor; confidence: medium; commits: 85611f002107, bc137951e966; files: src/plugins/gateway-startup-plugin-ids.ts, src/plugins/channel-plugin-ids.test.ts)
    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.
  2. added
    P1High-priority user-facing bug, regression, or broken workflow.
    impact:message-lossChannel message delivery can be lost, duplicated, or misrouted.
    impact:otherThis issue has meaningful maintainer-visible impact outside the owned taxonomy.
    issue-rating: 🦞 diamond lobsterVery strong issue quality with high-confidence source-level or clear reproduction.
    clawsweeper:source-reproClawSweeper found a high-confidence source-level issue reproduction.
    clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.
    clawsweeper:fix-shape-clearClawSweeper found a clear likely implementation shape for this issue.
    clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.
    clawsweeper:needs-product-decisionClawSweeper marked this issue as needing a product or behavior decision.
    on Jun 10, 2026
  3. added
    clawsweeper:linked-pr-openClawSweeper found an open linked pull request for this issue.
    clawsweeper:fix-shape-clearClawSweeper found a clear likely implementation shape for this issue.
    clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.
    and removed
    clawsweeper:fix-shape-clearClawSweeper found a clear likely implementation shape for this issue.
    clawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.
    clawsweeper:linked-pr-openClawSweeper found an open linked pull request for this issue.
    on Jun 19, 2026
  4. self-assigned this
    on Jul 6, 2026
  5. steipete commented on Jul 6, 2026

    @steipete
    Contributor

    Fixed on current main by #96397 / 16b275a.

    The Gateway now checks configured channel ownership after startup and emits a warning when an external Slack plugin is installed but blocked by explicit trust. The warning names channels.slack and gives the supported repair: enable plugins.entries.slack.enabled=true (and include slack in plugins.allow when that restrictive allowlist is configured). The merged PR includes focused startup tests plus real Gateway proof for both blocked and trusted channel plugins.

    We are keeping the fail-closed external-plugin trust boundary; the supported alternative to implicit loading is explicit enablement with a visible startup diagnostic. The current latest release, v2026.6.11, predates this merge, so the warning will appear in the next release containing the commit above.

    Closing as fixed on main. Thanks for the clear upgrade report—it directly motivated the startup diagnostic.

  6. locked as resolved and limited conversation to collaborators on Aug 2, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

P1High-priority user-facing bug, regression, or broken workflow.bugSomething isn't workingclawsweeper:needs-maintainer-reviewClawSweeper marked this issue as needing maintainer review before automation.clawsweeper:needs-product-decisionClawSweeper marked this issue as needing a product or behavior decision.clawsweeper:no-new-fix-prClawSweeper does not recommend queueing a new automated fix PR for this issue.clawsweeper:source-reproClawSweeper found a high-confidence source-level issue reproduction.impact:message-lossChannel message delivery can be lost, duplicated, or misrouted.impact:otherThis issue has meaningful maintainer-visible impact outside the owned taxonomy.issue-rating: 🦞 diamond lobsterVery strong issue quality with high-confidence source-level or clear reproduction.regressionBehavior that previously worked and now fails

Type

No type

Fields

Priority

None yet

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions