Back to skill

Security audit

Six Dim Evaluator

Security checks for vulnerabilities and agentic risk

Overview

The skill is not clearly malicious, but it asks for broad write and command execution authority while describing log analysis and database retention without enough scoping or user control.

Install only after reviewing whether you are comfortable granting write and shell execution powers to an evaluator. Prefer limiting it to read-only use unless you explicitly want it to run tests or generate files, avoid pointing it at private logs, and ask the publisher to document storage, retention, deletion, and dependency-update practices.

Vulnerability Patterns
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • 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)

T05 · Unauthorized Access and Privilege Escalation

Warning
Location
SKILL.md:10
Finding

Excessive Tool Permissions Violate Least-Privilege Requirements

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:10-14
Vulnerability Type: Excessive tool permissions
Risk Level: Medium

Vulnerable Code

yaml
tools:
  - Read
  - Write
  - Bash
  - Exec

Technical Analysis

The skill requests Write, Bash, and Exec capabilities even though the reviewed implementation only performs filesystem existence checks and reads files through Node.js fs APIs. No implemented feature requires writing files or executing operating-system commands.

This violates the principle of least privilege. In an agent environment where declared tools are granted to the loaded skill, a malicious or attacker-controlled evaluation context could attempt to influence the agent into using these unnecessary capabilities. The finding concerns the capability boundary exposed by the skill manifest; the reviewed source code does not itself invoke shell commands or write files.

Attack Path

  1. An agent loads the skill and grants the tools declared in SKILL.md.
  2. The skill receives Write, Bash, and Exec access despite only requiring read access for its implemented behavior.
  3. Attacker-controlled content or context encountered during an evaluation attempts to influence the agent's subsequent tool use.
  4. If the agent follows those instructions without sufficient independent authorization controls, it may execute commands or modify files using the unnecessary privileges.
  5. The resulting effects occur with the filesystem and operating-system permissions of the agent process.

Impact Assessment

Successful exploitation could permit arbitrary command execution or filesystem modification within the privileges of the hosting agent. Potential effects include alteration or deletion of accessible files, modification of project artifacts, access to data available to executed processes, and execution of additional local programs.

The practical scope depends on the host's to ...[truncated 221 chars]

Remediation
View remediation

Remediation Suggestions

  1. Restrict the declared tools to the minimum capability required by the current implementation:

    yaml
    tools:
      - Read
    
  2. Do not add Write, Bash, or Exec unless a concrete feature requires them.

  3. If future functionality must execute tests or generate reports, separate those operations into narrowly scoped capabilities with:

    • Explicit user confirmation before execution.
    • Command and argument allowlists.
    • Fixed working directories and canonicalized path validation.
    • Filesystem write restrictions limited to a dedicated output directory.
    • Process timeouts and resource limits.
    • A sandbox with no unnecessary credentials or network access.
  4. Add automated checks that compare declared permissions against actual implementation requirements and reject unnecessary privileged tools during review.

Vulnerability Patterns
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Behavioral ASTexec() Call, eval() Call, Dynamic Import
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (16)

Known Vulnerable Dependency: brace-expansion==2.0.3 — 3 advisory(ies): CVE-2026-13149 (brace-expansion: DoS via exponential-time expansion of consecutive non-expanding); CVE-2026-14257 (brace-expansion: DoS via unbounded expansion length causing an out-of-memory pro); CVE-2026-69152 (brace-expansion: DoS via unbounded intermediate arrays, bypassing the CVE-2026-1)

High
Category
Supply Chain
Confidence
95% confidence
Finding

brace-expansion 2.0.3 is reported with multiple denial-of-service issues involving pathological expansion patterns that can cause extreme CPU or memory consumption. Even though this copy is dev-only, it sits in common pattern-matching/globbing paths, so attacker-controlled patterns in CI, test discovery, or developer tooling could hang or exhaust resources.

Content

No source excerpt is available for this finding.

Known Vulnerable Dependency: browserslist==4.28.2 — 2 advisory(ies): CVE-2026-73088 (Browserslist: Uncaught crash / prototype write via untrusted browserslist-stats.); CVE-2026-73089 (Browserslist: Unbounded memory growth (no cache eviction) via distinct query res)

High
Category
Supply Chain
Confidence
90% confidence
Finding

browserslist 4.28.2 is flagged for crash/prototype-write and unbounded memory growth issues when handling untrusted stats or distinct query inputs. In this project it is a transitive dev dependency, which limits exposure, but any build/test workflow that ingests attacker-controlled browserslist queries or stats files could be destabilized or corrupted.

Content

No source excerpt is available for this finding.

Known Vulnerable Dependency: js-yaml==3.14.2 — 4 advisory(ies): CVE-2026-84375 (js-yaml: maxTotalMergeKeys does not limit CPU use for empty merge sources); CVE-2026-59869 (js-yaml: YAML merge-key chains can force quadratic CPU consumption); GHSA-5p4m-2wfm-xmqj (JS-YAML: Quadratic CPU consumption in !!omap resolution (3.x and 4.x) — CVE-2026) +1 more

High
Category
Supply Chain
Confidence
93% confidence
Finding

js-yaml 3.14.2 has multiple reported CPU-consumption issues triggered by crafted YAML constructs such as merge-key chains or !!omap resolution. Although dev-only here, YAML parsers are commonly exposed to configuration input, so if any test or tooling path parses attacker-supplied YAML, it could cause significant resource exhaustion and denial of service.

Content

No source excerpt is available for this finding.

External Model or Provider Selection

Medium
Category
Excessive Agency
Confidence
90% confidence
Finding

Skill selects an external model or provider that may use a different account or billing plan than the operator expects. Undisclosed model switches can cause unexpected cost or quota consumption.

Content

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

md
- Write
  - Bash
  - Exec
model: sonnet
---

# Six-Dim Evaluator - 六维评估器

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The skill explicitly says it will analyze usage logs as part of evaluation, which can involve user or system-derived data, but it does not provide an upfront privacy notice, scope limitation, or consent mechanism. In a skill with Read/Write/Bash/Exec capabilities, undocumented log analysis increases the risk of collecting sensitive operational data without user awareness.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The FAQ states that evaluation data will be saved to a database for trend analysis, but there is no prominent prior disclosure of retention, access controls, or opt-out behavior. Persistent storage of evaluation outputs may include metadata, usage patterns, or potentially sensitive assessment artifacts, creating privacy and compliance risks if users are not warned in advance.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The documentation scoring only rewards README content containing specific Chinese section names such as '使用场景' and '快速开始'. This creates a language-policy issue because skills documented in other languages are penalized by design, and the file does not provide user opt-in, locale selection, or a documented region-specific justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The code quality score increases only when SKILL.md contains Chinese keywords like '版本', '定位', and '核心功能'. That natural-language requirement forces a specific language/locale for higher scores and does not provide an opt-in or justified regional constraint.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The manifest describes an automated six-dimension evaluation engine that generates assessment reports and improvement suggestions. In code, many sub-dimension inputs are fixed constants such as architecture, performance, framework, insight, API design, self-monitoring, adoption, and error handling, so the reported evaluation is largely synthetic rather than actually executed from the target skill's artifacts.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The description and the rest of the document consistently enforce Chinese-language interaction and examples, which can constitute a locale policy issue when no user opt-in or language choice is provided. There is no statement that the skill is region-specific or that other languages are supported on request.

Content

No source excerpt is available for this finding.

Known Vulnerable Dependency: @babel/core==7.29.0 — 1 advisory(ies): CVE-2026-49356 (@babel/core: Arbitrary File Read via sourceMappingURL Comment)

Low
Category
Supply Chain
Confidence
85% confidence
Finding

The lockfile pins @babel/core 7.29.0, which the scanner reports as vulnerable to arbitrary file read via sourceMappingURL processing. In this skill, @babel/core is only a transitive development dependency through Jest, so exploitability is reduced, but if tests or tooling process attacker-controlled JavaScript during CI or local evaluation, the vulnerable parser behavior could still expose local files.

Content

No source excerpt is available for this finding.

Known Vulnerable Dependency: baseline-browser-mapping==2.10.16 — 1 advisory(ies): CVE-2026-45819 (baseline-browser-mapping process termination on invalid input causes denial of s)

Low
Category
Supply Chain
Confidence
81% confidence
Finding

baseline-browser-mapping 2.10.16 is flagged for a process-termination denial-of-service issue on invalid input. It is a dev-only transitive dependency under browserslist, so this is mainly relevant to build/test tooling rather than runtime skill behavior, but malformed input to tooling could still crash CI or local execution.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The description string explicitly presents the skill as "评估层 - 六维评估引擎" in Chinese, while the rest of the package metadata is otherwise English-facing. For a policy that disallows forcing a specific language without opt-in, this is a natural-language locale choice that is neither optional nor documented as region-specific.

Content

No source excerpt is available for this finding.

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
88% confidence
Finding

The runtime dependency uses a caret range (^11.0.0), which allows automatic adoption of future compatible releases. This can increase supply-chain risk because a compromised or breaking upstream release could be pulled in without explicit review, especially for a skill executed in automated environments.

Content

Scanner excerpt · package.json (reported line 25)May include surrounding context.

json
"tags": ["evaluation", "assessment", "six-dimensions", "meta-skill"]
  },
  "dependencies": {
    "commander": "^11.0.0"
  },
  "devDependencies": {
    "jest": "^30.3.0"

Unpinned Dependencies

Low
Category
Supply Chain
Confidence
81% confidence
Finding

The devDependency jest is also version-ranged with a caret, which can introduce supply-chain exposure during testing or CI workflows. While this is less severe than a runtime dependency, compromised developer-tooling packages can still affect build pipelines, test runners, or contributor machines.

Content

Scanner excerpt · package.json (reported line 28)May include surrounding context.

json
"commander": "^11.0.0"
  },
  "devDependencies": {
    "jest": "^30.3.0"
  }
}

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
93% confidence
Finding

The data structure includes hasClawHubListing and clawHubStats, and the M-dimension later uses hasClawHubListing to score marketplace presence. However, despite the stated intent to check for a ClawHub listing, no code actually populates these fields, so the documented behavior implies a lookup that never happens.

Content

No source excerpt is available for this finding.