ClawHub
Back to skill

Security audit

PR-Proof-of-Work

Security checks for vulnerabilities and agentic risk

Overview

This skill openly supports a GitHub PR workflow that captures and publishes E2E screenshots, but users should review screenshots and git actions before running it.

Install only if you want an agent-assisted GitHub PR workflow that can push branches and post PR comments under your authenticated GitHub account. Before running it, confirm the target repo, branch, and PR number, review all commits, and inspect screenshots for secrets or private data.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
95% confidence
Finding
The script does more than merely assemble and post a PR comment: it creates/updates a branch, commits files, and pushes them to the remote. That expands the trust boundary from local formatting to repository mutation and network publication, which can leak screenshots or alter repository state unexpectedly if invoked in automation.

Context-Inappropriate Capability

Medium
Confidence
97% confidence
Finding
Force-pushing a branch is a high-impact repository write capability that is not inherently necessary for generating a PR comment. Even if intended for temporary screenshot hosting, it can overwrite prior branch contents and publish potentially sensitive screenshots to a remote without meaningful safeguards.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill instructs the agent to create branches, commit, push to a remote, create PRs, and post screenshot artifacts without any explicit confirmation gate or warning that it will modify local git state and publish data to GitHub. In an automated or semi-automated setting, this can lead to unintended code publication, leakage of screenshots or repository contents, and irreversible repository changes if the wrong repo/branch is targeted.

Missing User Warnings

Medium
Confidence
90% confidence
Finding
The script pushes to a remote repository without any runtime warning, dry-run output, or confirmation, so a user may believe it only comments on a PR when it is actually publishing files and altering remote state. In CI or agent-driven contexts, this increases the chance of unintended disclosure and hard-to-audit repository changes.

Static analysis

No suspicious patterns detected.