ClawHub
Back to skill

Security audit

Fleet Conductor

Security checks for vulnerabilities and agentic risk

Overview

This skill coordinates repository changes and pull request merges with clear scope limits, review gates, and protections for unrelated work.

Use this skill only for repositories you authorize it to change and merge. Confirm that both review providers may receive the repository's code and issue content, especially for private projects. Expect parallel workers and retained campaign records.

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
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (2)

Mandatory AI reviews lack a warning about sharing repository content with external model providers.

Medium
Category
Output Handling
Confidence
92% confidence
Finding

The skill mandates reviews by model providers and directs reviewers to inspect repository diffs, issues, and instructions without first checking whether those materials may be shared with those providers. For private or sensitive repositories, this can disclose proprietary code or confidential information outside approved boundaries; read-only sessions and prompt-injection safeguards do not prevent that disclosure.

Content

No source excerpt is available for this finding.

Autonomous Decision Making

Medium
Category
Excessive Agency
Confidence
75% confidence
Finding

Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.

Content

Scanner excerpt · SKILL.md (reported line 160)May include surrounding context.

md
## Common mistakes

- Dispatching one lane per issue without checking shared write sets or contracts.
- Treating a worker's green report as GitHub check evidence.
- Releasing a held lane when its sibling opens a draft instead of when the recorded merge trigger fires.
- Marking a draft ready before reviewing its actual diff, receipts, collisions, and both independent reviews.