openclaw - ✅(Solved) Fix [Skills] 37 "Skipping skill path that resolves outside its configured root" warnings from built-in extensions [4 pull requests, 2 comments, 3 participants]

mashuai333 · 2026-03-12T06:19:35Z

[openclaw] markdown PR 44022: fix: stop false-positive built-in extension skill root warnings - Repository: openclaw/openclaw - Author: stim64045-spec - State:… ```markdown # PR #44022: fix: stop false-positive built-in extension skill root warnings - Repository: openclaw/openclaw - Author: stim64045-spec - State: closed | merged: False - Link: https://github.com/openclaw/openclaw/pull/44022 ## Description (problem / solution / changelog) ## Summary - make skill root containment fall back to device/inode ancestry when realpath strings diverge - keep the existing path escape rejection for true escapes - add a regression test for plugin skills loaded through aliased realpath ancestry ## Testing - pnpm exec vitest run src/agents/skills.loadworkspaceskillentries.test.ts src/agents/skills/plugin-skills.test.ts Closes #43745 ## Changed files - `src/agents/skills.loadworkspaceskillentries.test.ts` (modified, +81/-1) - `src/agents/skills/workspace.ts` (modified, +44/-1) --- # PR #16: Fix/acpx runtime skill copy - Repository: AcorpBG/openclaw - Author: AcorpBG - State: closed | merged: False - Link: https://github.com/AcorpBG/openclaw/pull/16 ## Description (problem / solution / changelog) ## Summary - Problem: `dist-runtime` staged bundled plugin skill trees as symlinks back into `dist/...`, so the skill loader's realpath containment check saw those bundled skills as escaping the configured runtime root and skipped them. - Why it matters: bundled extension skills such as `acpx/skills/acp-router` could disappear from the runtime overlay even though the containment guard was working as intended. - What changed: the runtime overlay still keeps manifest-declared bundled skill trees physically copied inside `dist-runtime`, but the manifest handling is now descriptor-driven and the Codex path resolution matches the canonical bundle semantics instead of additively copying the default `skills/` tree when a manifest declares a different root. - What did NOT change (scope boundary): this does not relax the skill-loader containment check, trust manifestless bundle layouts, or change unrelated symlink behavior outside the declared runtime skill trees. ## Change Type (select all) - [x] Bug fix - [ ] Feature - [ ] Refactor - [ ] Docs - [ ] Security hardening - [ ] Chore/infra ## Scope (select all touched areas) - [ ] Gateway / orchestration - [x] Skills / tool execution - [ ] Auth / tokens - [ ] Memory / storage - [x] Integrations - [ ] API / contracts - [ ] UI / DX - [x] CI/CD / infra ## Linked Issue/PR - Closes # - Related #43745 - Related #41651 - Related #44522 - Related #47601 - Related #48595 ## User-visible / Behavior Changes - Bundled extension skills declared by plugin manifests stay loadable from `dist-runtime` instead of being skipped as escaped paths. - Runtime staging now keeps its copy set aligned with the manifest-declared skill roots instead of widening Codex bundles to an extra default `skills/` tree when a custom root is declared. - Unrelated runtime-overlay artifacts keep the prior symlink/wrapper behavior. ## Security Impact (required) - New permissions/capabilities? (`Yes/No`) No - Secrets/tokens handling changed? (`Yes/No`) No - New/changed network calls? (`Yes/No`) No - Command/tool execution surface changed? (`Yes/No`) No - Data access scope changed? (`Yes/No`) No - If any `Yes`, explain risk + mitigation: Notes: - This remains a packaging/runtime-overlay fix, not a skill-loader security weakening. - The realpath containment check stays in place. - Manifest-declared paths still go through plugin-root containment checks before being copied. ## Repro + Verification ### Environment - OS: Ubuntu 24.04 - Runtime/container: `dist/` + `dist-runtime` bundled plugin overlay - Model/provider: N/A - Integration/channel (if any): : - Relevant config (redacted): ### Steps 1. Build/stage bundled plugins so the runtime overlay exposes bundled skill trees under `dist-runtime/extensions/ `. 2. Run the targeted runtime-staging test suite. 3. Let the skill loader resolve those paths with `realpath` against the configured bundled root under `dist-runtime`. ### Expected - Manifest-declared bundled skills remain physically inside `dist-runtime` and pass the existing containment check. - The runtime copy set stays limited to the declared skill roots for each supported manifest format. ### Actual - Before the patch, those skill paths were symlinked back into `dist/...`, so the loader resolved them outside the configured `dist-runtime` root and skipped them. - Before this review refinement, Codex bundles with an explicit custom skill root would also copy the default `skills/` tree additively inside `dist-runtime`, which was broader than the canonical bundle semantics. - After the final patch, declared bundled skill trees stay copied inside `dist-runtime`, and the Codex copy behavior stays bound to the declared roots. ## Evidence Attach at least one: - [x] Failing test/log before + passing after - [ ] Trace/log snippets - [

openclaw2026-03-12 06:19:35

ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

GitHub issue URL

Helpful · Quick feedback

GitHub stats

openclaw/openclaw#43745•Fetched 2026-04-08 00:17:34

View on GitHub

Comments

Participants

Timeline

Reactions

Author

Participants

Timeline (top)

cross-referenced ×5commented ×2labeled ×2subscribed ×2

Root Cause

Code Example

## 🐛 Bug Description

When running `openclaw doctor` or `openclaw skills list`, I get 37 warnings:

---

### Steps to reproduce

These warnings come from built-in extension skill directories, not from workspace skills.

## 🔍 Investigation Results

### Warning Sources

The warnings come from these built-in extensions:
- `~/.npm-global/lib/node_modules/openclaw/extensions/diffs/`
- `~/.npm-global/lib/node_modules/openclaw/extensions/feishu/`
- `~/.npm-global/lib/node_modules/openclaw/extensions/open-prose/`
- Other extension skill directories

### What I've Tried

1. ✅ **Workspace skills fixed**: Added `SKILL.md` to all auxiliary directories (scripts/, references/, assets/, etc.)
2. ✅ **Memory cleaned**: Removed duplicate entries from MEMORY.md
3. ✅ **No symlinks**: Verified no symbolic links in skill directories
4. ✅ **Realpath check**: All directories resolve correctly within their roots

### Test Results

---

**Conclusion**: Warnings come from `extensions/` directory skill loading.


### Expected behavior

No warnings should appear for built-in extension skills that are properly configured.

### Actual behavior

## 🔧 Impact

- **Functionality**: Skills work correctly, no functional impact
- **User Experience**: Warning spam in logs (37 warnings per command)
- **Performance**: Minimal (just log output)

## 📝 Suggestions

1. Fix the root path validation for extension skills
2. Add `.skillignore` support for auxiliary directories
3. Reduce warning verbosity for known patterns

### OpenClaw version

2026.3.8 (3caab92)

### Operating system

Linux 6.6.87.2-microsoft-standard-WSL2 (x64) - WSL2 on Windows

### Install method

npm global install

### Model

bailian/qwen3.5-plus

### Provider / routing chain

openclaw -> bailian

### Config file / key location

~/.openclaw/openclaw.json ; models.providers.bailian

### Additional provider/model setup details

Default model: bailian/qwen3.5-plus
Model pool configured with 8 models (qwen3.5-plus, qwen3-max, glm-5, etc.)

### Logs, screenshots, and evidence

RAW_BUFFERClick to expand / collapse

Bug type

Behavior bug (incorrect output/state without crash)

Summary

## 🐛 Bug Description

When running `openclaw doctor` or `openclaw skills list`, I get 37 warnings:

[skills] Skipping skill path that resolves outside its configured root.


### Steps to reproduce

These warnings come from built-in extension skill directories, not from workspace skills.

## 🔍 Investigation Results

### Warning Sources

The warnings come from these built-in extensions:
- `~/.npm-global/lib/node_modules/openclaw/extensions/diffs/`
- `~/.npm-global/lib/node_modules/openclaw/extensions/feishu/`
- `~/.npm-global/lib/node_modules/openclaw/extensions/open-prose/`
- Other extension skill directories

### What I've Tried

1. ✅ **Workspace skills fixed**: Added `SKILL.md` to all auxiliary directories (scripts/, references/, assets/, etc.)
2. ✅ **Memory cleaned**: Removed duplicate entries from MEMORY.md
3. ✅ **No symlinks**: Verified no symbolic links in skill directories
4. ✅ **Realpath check**: All directories resolve correctly within their roots

### Test Results

```bash
# Test 1: Rename workspace skills - warnings still appear
mv ~/.openclaw/workspace/skills ~/.openclaw/workspace/skills.bak
openclaw skills list | grep "Skipping skill"  # Still 37 warnings

# Test 2: Rename built-in skills - warnings still appear  
mv ~/.npm-global/lib/node_modules/openclaw/skills ~/.npm-global/lib/node_modules/openclaw/skills.bak
openclaw skills list | grep "Skipping skill"  # Still 37 warnings

# Test 3: Rename extensions - warnings disappear!
mv ~/.npm-global/lib/node_modules/openclaw/extensions ~/.npm-global/lib/node_modules/openclaw/extensions.bak
openclaw skills list | grep "Skipping skill"  # 0 warnings!

Conclusion: Warnings come from extensions/ directory skill loading.

Expected behavior

No warnings should appear for built-in extension skills that are properly configured.

Actual behavior

🔧 Impact

Functionality: Skills work correctly, no functional impact
User Experience: Warning spam in logs (37 warnings per command)
Performance: Minimal (just log output)

📝 Suggestions

Fix the root path validation for extension skills
Add .skillignore support for auxiliary directories
Reduce warning verbosity for known patterns

OpenClaw version

2026.3.8 (3caab92)

Operating system

Linux 6.6.87.2-microsoft-standard-WSL2 (x64) - WSL2 on Windows

Install method

npm global install

Model

bailian/qwen3.5-plus

Provider / routing chain

openclaw -> bailian

Config file / key location

~/.openclaw/openclaw.json ; models.providers.bailian

Additional provider/model setup details

Default model: bailian/qwen3.5-plus Model pool configured with 8 models (qwen3.5-plus, qwen3-max, glm-5, etc.)

Logs, screenshots, and evidence

Impact and severity

No response

Additional information

No response

extent analysis

Fix Plan

To resolve the issue of warnings appearing for built-in extension skills, we need to adjust the root path validation for these skills. Here are the steps:

Update the openclaw code to correctly handle the root path for extension skills.
Add support for a .skillignore file in auxiliary directories to ignore certain paths.

Example code changes:

// Update the skill path validation function
function validateSkillPath(skillPath, rootPath) {
  // Check if the skill path is within the root path
  if (skillPath.startsWith(rootPath)) {
    return true;
  }
  // Check if the skill path is in an auxiliary directory
  const auxDirs = ['.scripts', '.references', '.assets'];
  for (const auxDir of auxDirs) {
    if (skillPath.includes(auxDir)) {
      // Ignore auxiliary directories
      return false;
    }
  }
  // If none of the above conditions are met, skip the skill path
  return false;
}

// Add support for .skillignore file
function loadSkills(rootPath) {
  const skillIgnorePath = path.join(rootPath, '.skillignore');
  if (fs.existsSync(skillIgnorePath)) {
    const ignorePatterns = fs.readFileSync(skillIgnorePath, 'utf8').split('\n');
    // Filter out skills that match the ignore patterns
    const skills = getSkills(rootPath).filter(skill => {
      for (const pattern of ignorePatterns) {
        if (skill.path.includes(pattern)) {
          return false;
        }
      }
      return true;
    });
    return skills;
  }
  // If no .skillignore file exists, load all skills
  return getSkills(rootPath);
}

Verification

To verify that the fix worked, run the following commands:

openclaw doctor
openclaw skills list

Check that there are no warnings about skipping skill paths.

Extra Tips

Make sure to update the openclaw version to the latest one after applying the fix.
If you encounter any issues with the .skillignore file, check the file format and ensure that it is correctly formatted.

Vote matrix · Quick signals

Works

Did the solution work? Tap to confirm.

Easy Fix

Was it a quick fix?

Time Saver

Did it save you time?

Blocking

Was it severely blocking?

Common Issue

Are others likely hitting this too?

Flaky / Intermittent

Is it intermittent?

Verified / Reproducible

Can you reproduce it reliably?

FAQ

Expected behavior

No warnings should appear for built-in extension skills that are properly configured.

#api #ssr #installation #tensor shape #autograd error #mixed precision #training loop #device allocation #model download #tokenizer error

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

openclaw - ✅(Solved) Fix [Skills] 37 "Skipping skill path that resolves outside its configured root" warnings from built-in extensions [4 pull requests, 2 comments, 3 participants]

Recommended Tools

GitHub issue graph ai analysis

Root Cause

Fix Action

Fixed

PR fix notes

PR #44022: fix: stop false-positive built-in extension skill root warnings

Description (problem / solution / changelog)

Summary

Testing

Changed files

PR #16: Fix/acpx runtime skill copy

Description (problem / solution / changelog)

Summary

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

User-visible / Behavior Changes

Security Impact (required)

Repro + Verification

Environment

Steps

Expected

Actual

Evidence

Human Verification (required)

Compatibility / Migration

Failure Recovery (if this breaks)

Risks and Mitigations

AI-Assisted Disclosure

Changed files

PR #49198: Fix/dist-runtime skill copy instead of symlink

Description (problem / solution / changelog)

Summary

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

User-visible / Behavior Changes

Security Impact (required)

Repro + Verification

Environment

Steps

Expected

Actual

Evidence

Human Verification (required)

Compatibility / Migration

Failure Recovery (if this breaks)

Risks and Mitigations

Scope / Non-goals

AI-Assisted Disclosure

Changed files

PR #49215: fix(plugins): stage bundled skills inside runtime root

Description (problem / solution / changelog)

Summary

Change Type (select all)

Scope (select all touched areas)

Linked Issue/PR

User-visible / Behavior Changes

Security Impact (required)

Repro + Verification

Environment

Steps

Expected

Actual

Evidence

Human Verification (required)

Review Conversations

Compatibility / Migration

Failure Recovery (if this breaks)

Risks and Mitigations

Changed files

Code Example

Bug type

Summary

Expected behavior

Actual behavior

🔧 Impact

📝 Suggestions