ClawHub
Back to skill

Security audit

vx-repo-contract

Security checks for vulnerabilities and agentic risk

Overview

The skill is mostly a repository-layout checklist, but it also tells agents to post GitHub issue comments after code work without that authority being clearly advertised in the skill description.

Install only if you want agents using this skill to enforce both repository layout rules and a delivery-tracking habit. Before use, decide whether agents may post GitHub issue comments automatically, or require explicit user approval for any public comment.

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
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The manifest describes a repository layout contract focused on root files, justfile/vx.toml structure, AGENTS.md, and root artifacts. Lines L210-L224 add operational process requirements about issue comments, PRs, CI, and delivery tracking, which are not repository-layout checks and expand the skill's behavior beyond the stated five layout rules.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The introduction promises five rules that each ship with a command deciding pass/fail in CI. However, L222-L224 define a requirement about ending a turn with exactly one issue comment, but do not provide a concrete command or repo-local check comparable to the earlier rules, contradicting the document's claim of mechanically enforceable standalone checks.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.