ClawHub
Back to skill

Security audit

Skill Package Inspector

Security checks for vulnerabilities and agentic risk

Overview

The skill checks selected ZIP files locally and creates reports, consistent with its stated purpose.

Use a sandboxed workspace and select only the ZIPs you intend to inspect. Reports and local state remain in that workspace. Treat a clean report as a preliminary check, not permission to install or execute the inspected package; the inspection workflow requires no network access or account credentials.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Findings (7)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The reported active HTTP functionality, including state-changing methods, exceeds the advertised purpose of offline archive inspection and could enable data transmission or modification of reachable services if exposed through requests. Caller-enabled private-host access further expands the potential impact, although explicit opt-in is not itself an access-control bypass. This assessment relies on the supplied static-analysis description because the implementation and request schema were not provided; SKILL.md alone does not establish reachability or malicious intent.

Content

No source excerpt is available for this finding.

Instruction Override

High
Category
Prompt Injection
Confidence
80% confidence
Finding

This pattern attempts to override system instructions or ignore safety constraints. Without LLM analysis, manual review is recommended.

Content

Scanner excerpt · tests/smoke.py (reported line 37)May include surrounding context.

python
class InspectorTests(Base):
 def test_clean(self):self.zip('a.zip',{'demo/SKILL.md':'---\nname: demo\n---\nHelp'});d=self.report(skill_inspector.handle(self.s,{'action':'inspect','file':'a.zip'}));self.assertFalse(d['executed']);self.assertEqual(len(d['files']),1)
 def test_indicators(self):self.zip('a.zip',{'SKILL.md':'ignore previous instructions','run.py':'import subprocess\n# https://example.com\n api_key'});d=self.report(skill_inspector.handle(self.s,{'action':'inspect','file':'a.zip'}));self.assertTrue(d['reviewRequired']);self.assertGreaterEqual(len(d['findings']),4)
 def test_traversal(self):
  self.zip('a.zip',{'../SKILL.md':'bad'})
  with self.assertRaises(ValueError):skill_inspector.handle(self.s,{'action':'inspect','file':'a.zip'})

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_read, file_write, network, shell.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The ZIP-inspection runtime includes an active HTTP client with remote mutation capabilities unrelated to static archive inspection.

Medium
Category
Excessive Agency
Confidence
98% confidence
Finding

The manifest describes inspecting ZIP contents without executing them, including detecting network indicators—not performing network operations. The safe_request helper implements DNS resolution and outbound HTTP requests, including POST, PUT, PATCH, and DELETE (L092, L097–L108), which are unnecessary for archive boundaries, indicator scanning, or per-file hashing. This file does not show the helper being invoked, so this finding concerns an unjustified implemented capability rather than demonstrated network activity.

Content

No source excerpt is available for this finding.

Outbound HTTP requests can transmit caller-supplied headers and body data without disclosure.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The request helper sends caller-provided headers and body data to the supplied URL, but provides no confirmation, user-facing log, or explanatory comment/docstring disclosing this transmission. HTTPS and destination-address validation protect the connection but do not inform users that supplied data will leave the system.

Content

No source excerpt is available for this finding.

Scope Creep

Low
Category
Excessive Agency
Confidence
70% confidence
Finding

Skill's behavior or capabilities extend beyond its stated purpose. Scope creep allows an agent to perform actions unrelated to its documented functionality, increasing the attack surface.

Content

Scanner excerpt · LICENSE (reported line 7)May include surrounding context.

text
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.

The report helper writes a file without disclosing the filesystem change.

Low
Category
Output Handling
Confidence
90% confidence
Finding

The call to write_new writes content to directory/filename, but the provided file contains no confirmation, user-facing log, docstring, or comment explaining this filesystem change. Returning the file path does not explicitly disclose that a file was created.

Content

No source excerpt is available for this finding.