Back to skill

Security audit

Eval Skill

Security checks for vulnerabilities and agentic risk

Overview

This skill is a legitimate evaluator, but it reads full skill directories and forwards untrusted file contents to child agents without enough isolation or scope controls.

Install only if you intend to use a Chinese-language skill evaluator and are comfortable with it reading the selected skill directory and sending full text contents to child evaluators. Avoid pointing it at directories containing secrets, private configs, credentials, or unrelated files; treat reports about adversarial skills cautiously because embedded prompt-injection text may influence the child evaluators.

Vulnerability Patterns
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
SKILL.md:13
Finding

Untrusted Skill Content Is Injected into Child-Agent Instructions Without Isolation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:13-31; references/evaluation-protocol.md:3-7
Vulnerability Type: Prompt injection through untrusted evaluation material
Risk Level: Medium

Vulnerable Snippets

English rendering of the relevant instructions from SKILL.md:13-31:

markdown
### Step 1: Read the target skill

Read all files in the target skill directory, excluding binary and non-text files:
- `SKILL.md`
- `scripts/*` (`.sh`, `.py`, `.js`, etc.)
- `references/*` (`.md`, etc.)
- Other text configuration files

### Step 2: Start three child agents in parallel

Use the evaluation protocol in `references/evaluation-protocol.md`, populate it with
the evaluated skill information and the complete contents of every file, and then
spawn three child agents at the same time using `mode="run"`.

Each child agent task must contain:
1. A role declaration
2. Evaluated skill information
3. All evaluation materials, including complete file contents
4. Evaluation criteria
5. Output-format requirements

English rendering of the relevant instructions from references/evaluation-protocol.md:3-7:

markdown
Every time a child agent is spawned for evaluation, send the following complete
content as its task.

## Evaluated Skill Information

Populate this section with the target skill name, version, brief description,
and the complete contents of every file.

Technical Analysis

The evaluated files are untrusted, attacker-controlled input. The workflow copies their complete contents directly into the instruction context sent to three child agents. It does not require strong data delimiters, tell the child agents to treat embedded instructions solely as evidence, prohibit compliance with directives found in evaluated files, or constrain unnecessary child-agent tools.

Consequently, a malicious target file can include prompt-injection instructions that compete with the e ...[truncated 2031 chars]

Remediation
View remediation

Remediation Suggestions

  1. Place evaluated content inside explicit, uniquely generated data delimiters and state that everything within those delimiters is untrusted evidence, not executable instructions.
  2. Add a mandatory evaluator rule that directives, role declarations, tool requests, and output-format requests found inside evaluated files must never be followed.
  3. Require evaluators to report detected prompt-injection language as a security finding.
  4. Run child agents with no tools unless a tool is strictly necessary. If tools are required, use an explicit allowlist and deny shell execution, network access, writes, and access outside an isolated workspace.
  5. Validate SCORE_SUMMARY independently against a strict schema: require exactly the expected keys, integer values from 1 through 10, one summary block, and no malformed or duplicate fields.
  6. Do not rely solely on model-produced marker blocks. Cross-check every score against the associated written rationale and reject inconsistent responses.
  7. Apply file-count, file-size, and total-context limits. Reject unsupported file types and clearly report omitted or truncated material.
  8. Consider evaluating suspicious files in separate contexts and synthesizing findings only after stripping embedded directives.
  9. Update both SKILL.md and references/evaluation-protocol.md so these protections are mandatory rather than optional guidance.
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 (4)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 16)May include surrounding context.

md
- `SKILL.md`

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill is described in very broad terms ('帮我评估一下这个 skill') and the body instructs reading the entire target skill directory and spawning multiple sub-agents, but it does not define clear scope, authorization, or invocation constraints. That increases the chance the evaluator is triggered on untrusted or overly sensitive inputs and causes over-collection of data from arbitrary skill files, which can amplify prompt-injection and data-exposure risks.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The skill metadata and instructions are written entirely in Chinese and imply Chinese-language interaction, but the file does not offer a language choice or state that the skill is intended only for a Chinese-language context. This can violate language/locale policy when users have not opted into that locale.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The file’s title and all output requirements are written in Chinese, and lines L58-L59 explicitly require Chinese dimension names in the JSON output. This effectively forces a specific language/locale for the skill evaluation protocol without any opt-in or explanation that it is intended only for a Chinese-language context.

Content

No source excerpt is available for this finding.