feat(skills): add clawsec-clawhub-checker reputation checking skill by abutbul · Pull Request #41 · prompt-security/clawsec

abutbul · 2026-02-16T17:48:12Z

Opener Type

Human
Agent (automated)

Summary

Adds a new skill that enhances the ClawSec guarded skill installer with ClawHub reputation checks and VirusTotal Code Insight integration. Provides a defense-in-depth layer for skill installation security.

Changes Made

New skill clawsec-clawhub-checker with reputation scoring (0-100) using multiple signals
Enhanced guarded installer with reputation checks before skill installation
VirusTotal Code Insight integration via ClawHub CLI
New exit code 43 for reputation warnings requiring --confirm-reputation
Integration with existing clawsec-advisory-guardian hook
Standalone skill compatible with dynamic catalog system

Related Issues

N/A - New feature contribution

Type of Change

Bug fix (non-breaking change that fixes an issue)
New feature (non-breaking change that adds functionality)
Breaking change (fix or feature that would cause existing functionality to change)
Documentation update
Security incident (please open a Security Incident Report issue instead of a PR)

Testing

Input validation tests (command injection prevention)
Argument parsing tests for enhanced installer
formatReputationWarning output formatting tests
All 11 new tests passing
All 24 existing tests still passing
ESLint clean, TypeScript clean, build clean

Run tests: node skills/clawsec-clawhub-checker/test/reputation_check.test.mjs

Checklist

My code follows the project's style guidelines
I have performed a self-review of my changes
I have added tests that prove my fix/feature works
New and existing tests pass locally

Limitation Note

This skill catches VirusTotal Code Insight flags but cannot access OpenClaw internal check results (only shown on ClawHub website, not exposed via CLI/API). Heuristic checks (skill age, author reputation, downloads) provide similar risk assessment.

- Adds ClawHub reputation checks to guarded installer - Integrates VirusTotal Code Insight scores - Requires --confirm-reputation for suspicious skills - Enhances advisory guardian hook with reputation warnings - Defense-in-depth layer for skill installation security

- Enhanced guarded installer with reputation checks - VirusTotal Code Insight integration - Reputation scoring (0-100) with multiple signals - New exit code 43 for reputation warnings - Requires --confirm-reputation for suspicious skills - Integration with clawsec-advisory-guardian hook - Standalone skill compatible with dynamic catalog system Note: Removed hardcoded catalog entry to work with new dynamic catalog system (discover_skill_catalog.mjs).

- Remove unused imports (fs, os, path) from check_clawhub_reputation.mjs - Remove unused variable in setup_reputation_hook.mjs - Remove unused os import from update_suite_catalog.mjs - All ESLint checks now pass - TypeScript check passes - Build check passes

davida-ps · 2026-02-16T18:25:42Z

PR Notes for ClawSec ClawHub Checker

Important Limitation Notice

This skill currently catches VirusTotal Code Insight flags but cannot access OpenClaw internal check results because:

VirusTotal flags are exposed via clawhub install command output (we parse stderr)
OpenClaw internal checks are only shown on the ClawHub website, not exposed via API

Example from `clawsec-suite` page:

✅ VirusTotal: "Benign"
⚠️ OpenClaw internal check: "The package is internally consistent with a feed-monitoring / advisory-guardian purpose, but a few operational details and optional bypasses deserve attention before installing."

Recommendation for ClawHub

Expose internal check results via:

clawhub inspect --json endpoint
Additional API field for security tools
Or at minimum, include in clawhub install warning output

Current Workaround

Our heuristic checks (skill age, author reputation, downloads, updates) provide similar risk assessment but miss specific operational warnings about bypasses, missing signatures, etc.

PR Should Include

This skill as defense-in-depth layer
Feature request to ClawHub for exposing internal check data
Documentation about the limitation

davida-ps · 2026-02-16T18:27:24Z

openclaw/clawhub#362

…d SKILL.md feat: add input validation for skill slug and version in check_clawhub_reputation.mjs fix: enhance argument parsing in enhanced_guarded_install.mjs test: add reputation check tests for input validation and output formatting chore: delete unused update_suite_catalog.mjs script

Copilot

Pull request overview

This PR adds a new ClawSec skill that enhances the guarded skill installer with ClawHub reputation checking capabilities. The skill integrates VirusTotal Code Insight scores and heuristic checks (skill age, author reputation, downloads) to provide defense-in-depth security before skill installation.

Changes:

New reputation checking skill with scoring algorithm (0-100) based on multiple security signals
Enhanced installer wrapper that performs reputation checks before delegating to original guarded installer
Setup script to integrate reputation checks into clawsec-advisory-guardian hook
New exit code 43 for reputation warnings requiring --confirm-reputation flag
Comprehensive test suite with 11 tests covering input validation and argument parsing

Reviewed changes

Copilot reviewed 8 out of 8 changed files in this pull request and generated 11 comments.

Show a summary per file

File	Description
`skills/clawsec-clawhub-checker/skill.json`	Skill metadata, dependencies, SBOM, and integration specifications
`skills/clawsec-clawhub-checker/scripts/check_clawhub_reputation.mjs`	Core reputation checking logic with input validation and VirusTotal integration
`skills/clawsec-clawhub-checker/scripts/enhanced_guarded_install.mjs`	Enhanced installer wrapper that adds reputation checks before skill installation
`skills/clawsec-clawhub-checker/scripts/setup_reputation_hook.mjs`	Integration script to modify clawsec-suite handler.ts with reputation checks
`skills/clawsec-clawhub-checker/hooks/clawsec-advisory-guardian/lib/reputation.mjs`	Reputation checking module for hook integration
`skills/clawsec-clawhub-checker/test/reputation_check.test.mjs`	Comprehensive test suite for reputation checking and input validation
`skills/clawsec-clawhub-checker/SKILL.md`	Main user-facing documentation with installation and usage instructions
`skills/clawsec-clawhub-checker/README.md`	Developer documentation with architecture and testing guidance

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

skills/clawsec-clawhub-checker/scripts/setup_reputation_hook.mjs

skills/clawsec-clawhub-checker/scripts/check_clawhub_reputation.mjs

skills/clawsec-clawhub-checker/SKILL.md

Copilot · 2026-02-16T18:46:31Z

skills/clawsec-clawhub-checker/scripts/enhanced_guarded_install.mjs

+async function main() {
+  try {
+    const args = parseArgs(process.argv.slice(2));
+
+    // Build args for original script (excluding reputation-specific args)
+    const originalArgs = [];
+    for (let i = 0; i < process.argv.slice(2).length; i++) {
+      const token = process.argv.slice(2)[i];
+      if (token === "--confirm-reputation" || token === "--reputation-threshold") {
+        i += token === "--reputation-threshold" ? 1 : 0;
+        continue;
+      }
+      originalArgs.push(token);
+    }
+
+    args.originalArgs = originalArgs;
+


The code parses arguments twice: once via parseArgs(process.argv.slice(2)) at line 130, and again manually at lines 133-141 to filter reputation-specific arguments. This duplication is error-prone and could lead to inconsistencies if argument parsing logic changes.

Consider refactoring to build originalArgs during the initial parseArgs call, or have parseArgs return both the parsed object and the filtered array for the original script. This would ensure a single source of truth for argument parsing logic.

Suggested change

async function main() {

try {

const args = parseArgs(process.argv.slice(2));

// Build args for original script (excluding reputation-specific args)

const originalArgs = [];

for (let i = 0; i < process.argv.slice(2).length; i++) {

const token = process.argv.slice(2)[i];

if (token === "--confirm-reputation" || token === "--reputation-threshold") {

i += token === "--reputation-threshold" ? 1 : 0;

continue;

}

originalArgs.push(token);

}

args.originalArgs = originalArgs;

function buildOriginalArgs(argv) {

const originalArgs = [];

for (let i = 0; i < argv.length; i++) {

const token = argv[i];

if (token === "--confirm-reputation" || token === "--reputation-threshold") {

if (token === "--reputation-threshold" && i + 1 < argv.length) {

// Skip the value associated with --reputation-threshold

i += 1;

}

continue;

}

originalArgs.push(token);

}

return originalArgs;

}

async function main() {

try {

const cliArgs = process.argv.slice(2);

const args = parseArgs(cliArgs);

// Build args for original script (excluding reputation-specific args)

args.originalArgs = buildOriginalArgs(cliArgs);

Copilot · 2026-02-16T18:46:31Z

skills/clawsec-clawhub-checker/skill.json

+      {
+        "path": "README.md",
+        "required": false,
+        "description": "Additional documentation and development guide"


The test file is not included in the SBOM. For consistency with the repository's testing practices and to ensure tests are distributed with the skill, consider adding the test directory to the SBOM:

{ "path": "test/reputation_check.test.mjs", "required": false, "description": "Test suite for reputation checking functionality" }

Suggested change

"description": "Additional documentation and development guide"

"description": "Additional documentation and development guide"

},

{

"path": "test/reputation_check.test.mjs",

"required": false,

"description": "Test suite for reputation checking functionality"

skills/clawsec-clawhub-checker/scripts/enhanced_guarded_install.mjs

skills/clawsec-clawhub-checker/SKILL.md

skills/clawsec-clawhub-checker/scripts/check_clawhub_reputation.mjs

skills/clawsec-clawhub-checker/scripts/enhanced_guarded_install.mjs

…n checks

…ce installer

…scripts and usage instructions

Copilot

Pull request overview

Copilot reviewed 8 out of 8 changed files in this pull request and generated 8 comments.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-02-16T19:12:11Z

skills/clawsec-clawhub-checker/scripts/check_clawhub_reputation.mjs

+// CLI interface for direct usage
+if (import.meta.url === `file://${process.argv[1]}`) {
+  async function main() {


The CLI entrypoint guard import.meta.url === file://${process.argv[1]}`` won’t work when the script is invoked via a relative path (e.g., node scripts/check_clawhub_reputation.mjs ... as shown in docs), because import.meta.url is absolute while argv[1] may be relative. Use a robust comparison (e.g., resolve argv[1] and convert to a file URL) or remove the guard.

Copilot · 2026-02-16T19:12:11Z

skills/clawsec-clawhub-checker/scripts/check_clawhub_reputation.mjs

+
+    const skillSlug = args[0];
+    const version = args[1] || "";
+    const threshold = args[2] ? parseInt(args[2], 10) : 70;


The CLI parsing allows an invalid threshold: parseInt can yield NaN (or an out-of-range number), which then makes result.safe = result.score >= threshold behave incorrectly (NaN makes it always false). Validate threshold (finite integer 0-100) and either default to 70 or exit 1 with a clear error.

Suggested change

const threshold = args[2] ? parseInt(args[2], 10) : 70;

let threshold;

if (args[2] === undefined) {

threshold = 70;

} else {

const parsedThreshold = parseInt(args[2], 10);

if (!Number.isInteger(parsedThreshold) || parsedThreshold < 0 || parsedThreshold > 100) {

console.error(

`Invalid threshold: "${args[2]}". Threshold must be an integer between 0 and 100.`

);

process.exit(1);

}

threshold = parsedThreshold;

}

Copilot · 2026-02-16T19:12:11Z

skills/clawsec-clawhub-checker/scripts/enhanced_guarded_install.mjs

+    if (token === "--version") {
+      parsed.version = String(argv[i + 1] ?? "").trim();
+      i += 1;
+      continue;


--version is accepted without any validation here. Because --confirm-reputation skips the call to checkClawhubReputation entirely, a user can bypass the checker’s version-format validation by passing --confirm-reputation. Consider always validating --version format in parseArgs, and only using --confirm-reputation to bypass the reputation threshold decision (not input validation).

Copilot · 2026-02-16T19:12:12Z

skills/clawsec-clawhub-checker/SKILL.md

+- `1` - Error
+


This section documents CLAWHUB_ALLOW_SUSPICIOUS and CLAWHUB_VIRUSTOTAL_API_KEY, but there is no implementation in this skill that reads or applies these environment variables. Either implement them or remove/clarify them so users don’t assume the behavior exists.

Suggested change

- `1` - Error

- `CLAWHUB_ALLOW_SUSPICIOUS` - Reserved for future use; not currently read by this skill. Intended to allow installation of suspicious skills without confirmation.

- `CLAWHUB_VIRUSTOTAL_API_KEY` - Reserved for future use; not currently read by this skill. Intended for using your own VirusTotal API key for deeper scans.

Copilot · 2026-02-16T19:12:12Z

skills/clawsec-clawhub-checker/README.md

+## Reputation Signals Checked
+
+1. **VirusTotal Code Insight** - Malicious code patterns
+2. **Skill Age** - New skills (<7 days) are riskier


README.md lists CLAWHUB_ALLOW_SUSPICIOUS as a supported environment variable, but none of the scripts read it. Either implement the behavior or remove it from the docs to avoid misleading configuration guidance.

Suggested change

2. **Skill Age** - New skills (<7 days) are riskier

Copilot · 2026-02-16T19:12:12Z

skills/clawsec-clawhub-checker/scripts/setup_reputation_hook.mjs

+      handlerContent = handlerContent.slice(0, lineEndIndex + 1) + `${importLine}\n` + handlerContent.slice(lineEndIndex + 1);
+      handlerChanged = true;
+    } else {
+      console.log("✓ Hook handler already imports reputation module");
+    }


The insertion logic searches for the exact line const matches = findMatches(feed, installedSkills);, but the current clawsec-suite hook handler uses const allMatches = findMatches(feed, installedSkills); and later declares matches separately. This means findMatchesLine will be -1 and the script will skip integration (only printing the warning). Update the search/insert to match the current handler.ts structure (or use AST-based editing).

Copilot · 2026-02-16T19:12:12Z

skills/clawsec-clawhub-checker/scripts/setup_reputation_hook.mjs

+    const reputationModuleDst = path.join(hookLibDir, "reputation.mjs");
+
+    await fs.copyFile(reputationModuleSrc, reputationModuleDst);
+    console.log(`✓ Copied reputation module to ${reputationModuleDst}`);


setup_reputation_hook.mjs copies reputation.mjs into clawsec-suite, but the copied module executes scripts/check_clawhub_reputation.mjs relative to its own location. Since clawsec-suite does not contain that script, the hook integration will fail at runtime unless you also copy check_clawhub_reputation.mjs into the suite (or change the module to locate the checker skill directory explicitly).

Suggested change

console.log(`✓ Copied reputation module to ${reputationModuleDst}`);

console.log(`✓ Copied reputation module to ${reputationModuleDst}`);

// Copy the ClawHub reputation checker script so reputation.mjs can resolve it relative to the suite root

const reputationScriptSrc = path.join(checkerDir, "scripts", "check_clawhub_reputation.mjs");

await fs.mkdir(suiteScriptsDir, { recursive: true });

const reputationScriptDst = path.join(suiteScriptsDir, "check_clawhub_reputation.mjs");

await fs.copyFile(reputationScriptSrc, reputationScriptDst);

console.log(`✓ Copied ClawHub reputation checker script to ${reputationScriptDst}`);

Copilot · 2026-02-16T19:12:13Z

skills/clawsec-clawhub-checker/hooks/clawsec-advisory-guardian/lib/reputation.mjs

+    const reputationCheck = spawnSync(
+      "node",
+      [
+        `${checkerDir}/scripts/check_clawhub_reputation.mjs`,
+        skillSlug,


This spawnSync call runs ${checkerDir}/scripts/check_clawhub_reputation.mjs where checkerDir is derived from the current file location. After setup_reputation_hook.mjs copies this module into clawsec-suite/hooks/..., checkerDir will resolve to the suite root (which does not contain that script), so the hook’s reputation check will fail. Make the script location resolution independent of the file’s location, or ensure the setup step copies the checker script into the suite.

…ld; update documentation

abutbul added 3 commits February 16, 2026 13:01

abutbul force-pushed the skill/clawsec-clawhub-checker branch from f70aa5e to e6b9e90 Compare February 16, 2026 17:53

Merge branch 'main' into skill/clawsec-clawhub-checker

269ff94

davida-ps changed the title ~~Skill/clawsec clawhub checker~~ feat(skills): add clawsec-clawhub-checker reputation checking skill Feb 16, 2026

davida-ps requested a review from Copilot February 16, 2026 18:37

Copilot started reviewing on behalf of davida-ps February 16, 2026 18:38 View session

Copilot AI reviewed Feb 16, 2026

View reviewed changes

feat: enhance clawsec-clawhub-checker with setup script and reputatio…

6893390

…n checks

davida-ps requested a review from Copilot February 16, 2026 19:01

Copilot started reviewing on behalf of davida-ps February 16, 2026 19:01 View session

davida-ps added 2 commits February 16, 2026 21:09

feat: integrate reputation checks into clawhub setup script and enhan…

1530958

…ce installer

docs: update README and SKILL documentation to reflect new installer …

4538622

…scripts and usage instructions

Copilot AI reviewed Feb 16, 2026

View reviewed changes

feat: enhance CLI validation for skill version and reputation thresho…

6d6bb6a

…ld; update documentation

davida-ps merged commit f43f792 into prompt-security:main Feb 16, 2026
13 checks passed

-    const threshold = args[2] ? parseInt(args[2], 10) : 70;
+    let threshold;
+    if (args[2] === undefined) {
+      threshold = 70;
+    } else {
+      const parsedThreshold = parseInt(args[2], 10);
+      if (!Number.isInteger(parsedThreshold) || parsedThreshold < 0 || parsedThreshold > 100) {
+        console.error(
+          `Invalid threshold: "${args[2]}". Threshold must be an integer between 0 and 100.`
+        );
+        process.exit(1);
+      }
+      threshold = parsedThreshold;
+    }

	- `1` - Error
	- `CLAWHUB_ALLOW_SUSPICIOUS` - Reserved for future use; not currently read by this skill. Intended to allow installation of suspicious skills without confirmation.
	- `CLAWHUB_VIRUSTOTAL_API_KEY` - Reserved for future use; not currently read by this skill. Intended for using your own VirusTotal API key for deeper scans.

-    console.log(`✓ Copied reputation module to ${reputationModuleDst}`);
+    console.log(`✓ Copied reputation module to ${reputationModuleDst}`);
+    // Copy the ClawHub reputation checker script so reputation.mjs can resolve it relative to the suite root
+    const reputationScriptSrc = path.join(checkerDir, "scripts", "check_clawhub_reputation.mjs");
+    await fs.mkdir(suiteScriptsDir, { recursive: true });
+    const reputationScriptDst = path.join(suiteScriptsDir, "check_clawhub_reputation.mjs");
+    await fs.copyFile(reputationScriptSrc, reputationScriptDst);
+    console.log(`✓ Copied ClawHub reputation checker script to ${reputationScriptDst}`);

Conversation

abutbul commented Feb 16, 2026 • edited by davida-ps Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Opener Type

Summary

Changes Made

Related Issues

Type of Change

Testing

Checklist

Limitation Note

Uh oh!

davida-ps commented Feb 16, 2026

PR Notes for ClawSec ClawHub Checker

Important Limitation Notice

Example from clawsec-suite page:

Recommendation for ClawHub

Current Workaround

PR Should Include

Uh oh!

davida-ps commented Feb 16, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI Feb 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 16, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Feb 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 16, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 16, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

abutbul commented Feb 16, 2026 •

edited by davida-ps

Loading

Example from `clawsec-suite` page: