Back to skill

Security audit

skill-review-pro

Security checks for vulnerabilities and agentic risk

Overview

The scanners report a benign skill with no AIG threat detections and several purpose-related cautions, though Skillspector coverage was partial and confidence is medium.

The verdict is benign, but consider the medium confidence and partial coverage before installing. Review the unpinned package references and the remote CI rerun option, and be aware that generic fix triggers could activate outside a skill review.

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
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (9)

MCP server referenced without pinned version: 'npx clawhub'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx clawhub'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

The publish-recovery workflow permits the agent to rerun remote CI jobs despite its declared read-only tool boundary.

Medium
Category
Excessive Agency
Confidence
94% confidence
Finding

L373 allows the agent to execute gh run rerun <run-id> --failed after explicit user authorization; L674 repeats the delegation option. This is a remote mutation using a command absent from the allowed-tools list at L005, and it contradicts the read-only Bash and publish-verification restrictions at L023–L025. User confirmation alone does not bring this operation within the skill's declared permissions.

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 · fix/SKILL.md (reported line 154)May include surrounding context.

md
## 约束 / Constraints

- **绝不主动修改** — 每条必须经用户确认 / Never modify without confirmation
- **只修清单里的内容** — 不擅自扩大范围 / Only fix what's in the checklist
- **中英双语同步(仅当该 skill 本身是双语时)** — 修中文必须同步修英文;单语 skill 只维护其原有语言 / If the target skill is bilingual, keep both languages in sync; monolingual skills keep their original language only
- **不动 frontmatter** — 除非清单明确指出 / Don't touch frontmatter unless specified

MCP server referenced without pinned version: 'npx --yes clawhub'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx --yes clawhub'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx --yes clawhub'.

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

subprocess module call

Medium
Category
Dangerous Code Execution
Confidence
70% confidence
Finding

subprocess module calls execute external commands. Without careful input validation, this enables command injection.

Content

Scanner excerpt · scripts/publish_check.py (reported line 68)May include surrounding context.

python
cmd = clawhub_bin.split() + ["inspect", slug, "--json"]
    try:
        # 安全说明:cmd 来自 clawhub_bin.split() + 固定参数列表,argv 直传、不经 shell、无拼接用户输入
        proc = subprocess.run(cmd, capture_output=True, text=True, timeout=timeout)
    except subprocess.TimeoutExpired:
        return None, f"inspect 超时(>{timeout}s)"
    if proc.returncode != 0:

Direct-fix triggers use generic words without an explicit Skill-review context requirement.

Low
Category
Trigger Abuse
Confidence
91% confidence
Finding

The mode explicitly activates on standalone words such as "improve" and "enhance", which commonly occur in requests unrelated to Skill review. Although the surrounding description discusses improving a Skill, the trigger rule does not require an identified target Skill or an active Skill-review conversation, allowing unintended routing into the fix workflow.

Content

No source excerpt is available for this finding.