src/node-host/runner.credentials.test.ts was removed in 4dff4e95b1
(#107270) but stayed in forcedUnitFastTestFiles, breaking
test/vitest-unit-fast-config.test.ts on every branch.
* fix(skills): link globally-installed ClawHub skills via managed lockfile
buildWorkspaceSkillStatus read the ClawHub lockfile only from the workspace
dir, so globally-installed skills (source "openclaw-managed", tracked in the
managed lockfile at <dirname(managedSkillsDir)>/.clawhub/lock.json) were
falsely reported as "not tracked by the workspace ClawHub lockfile". Resolve
openclaw-managed skills against the managed lockfile and managed parent dir
so the link status and expected-install-dir check match. Workspace-local
skills are unchanged.
Closes#105530
Co-Authored-By: Claude <noreply@anthropic.com>
* fix(skills): make ClawHub lockfile diagnostics scope-aware
resolveClawHubSkillStatusLinkSync hardcoded "workspace ClawHub lockfile" in
its invalid-path diagnostics, so a globally-managed skill (checked against
the managed lockfile) still got a misleading "workspace" troubleshooting
message. Add an optional lockfileScope ("workspace" | "managed") and use it
in the diagnostics; buildSkillStatus passes "managed" for openclaw-managed
skills. Extends the managed missing-entry regression test to assert the
managed diagnostic.
Addresses ClawSweeper [P2] review finding on #106479.
Co-Authored-By: Claude <noreply@anthropic.com>
* fix(skills): relabel managed-lockfile diagnostics from status.ts
Compute the managed-lockfile diagnostic relabel in status.ts instead of
threading a scope parameter through clawhub.ts. clawhub.ts is an oversized
legacy module under the check:loc ratchet ('oversized legacy files may not
grow'), so the prior +8 net lines tripped checks-fast-loc-ratchet. Behavior
is unchanged: globally-managed skills still surface 'managed ClawHub
lockfile' diagnostics.
* fix(skills): strengthen managed ClawHub status linkage
---------
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
* refactor(channels): migrate irc, nextcloud-talk, zalouser, slack, and matrix group policy onto the scope tree
* fix(irc): drop unused tree binding in group match
* fix(zalouser): keep wildcard lookup opt-in for explicit-only candidates
* fix(zalouser): adapter opts into the wildcard fallback candidate
* refactor(channels): migrate feishu, msteams, and discord group policy onto the scope tree and drop the core discord duplicate
* fix(msteams): restore the cross-team scan fallback for policy-less matched teams
* fix(feishu): keep the adapter's zod-typed group config for the scope tree
* revert(core): keep the discord require-mention fallback until stage 3 makes it provably redundant
* chore(matrix): drop the now-unused channel entry match re-export
* refactor(telegram): migrate group policy onto the scope tree
* test(telegram): keep the bot token fixture off secret-scanner patterns
* test(telegram): use a computed key for the bot token fixture
* test(telegram): assemble the bot token fixture indirectly for scanner and lint
* fix(extra-params): preserve resolved cacheRetention when options carries undefined own-property
When the proxy transport emits cacheRetention as an own property set to
undefined, JS spread semantics clobber the resolved cacheRetention value
from per-model params. Re-assert the resolved value after the spread so
it takes precedence over an undefined own-property in the caller's options.
Fixes#106014
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
* test(agents): strengthen cache retention precedence proof
* refactor(agents): keep cache retention merge compact
* style(agents): format cache retention matrix
---------
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
Co-authored-by: Peter Steinberger <steipete@gmail.com>
Adds consistent manual Workboard movement across CLI, slash commands, and agent tools while preserving operator authority and agent claim isolation.
Co-authored-by: Ross I4JU <ross@i4ju.ch>
