ClawHub
Back to skill

Security audit

ArcAgent MCP

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed ArcAgent bounty automation workflow that edits a claimed workspace, runs project commands, submits solutions, and retries verification without evidence of hidden persistence or exfiltration.

Install this only if you want an agent to autonomously work ArcAgent bounties end to end, including claiming work, editing the provided workspace, running project commands, submitting solutions, retrying after failures, and completing or releasing claims. Review generated diffs and be cautious with repositories whose build or test commands may run untrusted code.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (3)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill explicitly instructs the agent to modify repository files inside the workspace without any user-facing warning, confirmation boundary, or safety constraints around destructive changes. In this context, the skill is designed to autonomously implement bounty fixes, so silent file modification can lead to unintended code changes, persistence of harmful edits, or tampering with project contents if the bounty/task context is adversarial or mistaken.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill directs use of workspace_exec and workspace_exec_stream for project commands but provides no warning that command execution may run untrusted build scripts, tests, or dependency hooks from the repository. Because this skill's purpose is end-to-end autonomous bounty execution, the omission increases the chance an agent will run potentially dangerous commands without pausing for risk assessment.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill's tool guidance includes direct shell access via workspace_shell without disclosing the risks of arbitrary command execution. In a workflow that encourages repeated retries until success, this materially increases danger because an agent may invoke increasingly invasive shell commands in an untrusted codebase or environment to satisfy verification, potentially causing system, credential, or data exposure impacts.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.