Back to skill

Security audit

胡田 OPC导师 迭代训练

Security checks for vulnerabilities and agentic risk

Overview

This skill is coherent for improving other skills, but it can broadly read and modify skill files through passive or scheduled triggers without clear user confirmation.

Install only if you want a skill that can maintain and modify other local skills. Before using it, require explicit user approval for each training run, specify the exact target skill path, review diffs before any write, and keep version-control or backups for rollback.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (5)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger words include very common conversational phrases such as '训练', '练一下', and '迭代', which can cause the skill to activate in situations where the user did not intend to invoke a file-writing training workflow. Because the skill later reads and updates local skill artifacts, ambiguous activation raises the chance of unintended execution and accidental workspace changes.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill can activate through several broad paths, including passive triggers from reports, dissatisfaction with output, and scheduled retraining, without defining strict user-consent boundaries. In a multi-tool agent environment, this creates a risk that the skill initiates analysis or file modification based on inferred state rather than an explicit user request.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly instructs updating SKILL.md or related reference documents as part of iteration, but it does not require warning the user before altering local files. This is dangerous because it normalizes silent prompt-driven modification of agent behavior definitions, which can unintentionally overwrite content or introduce unsafe changes into reusable skills.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The tool/data section states that the skill will read and modify files under './skills/{Skill名}/', but it does not present any access constraints, confirmation requirement, or least-privilege boundaries. This increases risk because a loosely triggered skill could traverse and alter operational prompt files that affect future agent behavior across the workspace.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
94% confidence
Finding

The natural-language content throughout the skill README is exclusively Chinese, including the title, description, and usage instructions. Under the policy, forcing a specific language without user opt-in can be a locale-policy violation when no alternative language option or justification is provided.

Content

No source excerpt is available for this finding.