Back to skill

Security audit

Token省钱管家

Security checks for vulnerabilities and agentic risk

Overview

This skill is a local token-cost helper with disclosed caching, routing, monitoring, and controls, though users should review its local data storage and skill-disabling controls before use.

Install this only if you are comfortable with a Chinese-language token cost tool that writes local JSON files for cache, cost history, alerts, and control settings. Avoid caching sensitive prompts or outputs unless the storage location and file permissions are acceptable, and review disabled-skill or budget settings because they can block workflows when the host honors these controls.

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
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (22)

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The declared description promises an active cost-optimization system with caching, routing, monitoring, and automatic runtime optimization. The supplied code chunk does not implement any of those core behaviors. Instead, it performs setup tasks: creating a local directory and writing a JSON configuration file for control/cost state. While installation/setup can support such a tool, this chunk alone materially differs from the declared primary purpose and also includes control fields like disabledSkills and pausedPlans that are not mentioned in the description. Therefore, the description does not accurately represent what this code chunk actually does.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
91% confidence
Finding

The supplied code only covers one supporting piece of the declared functionality: caching. It provides cache key generation, TTL-based retrieval, bounded storage, simple eviction, hit-rate statistics, and disk persistence via fs/path. However, the declared description emphasizes a broader cost-optimization system with dynamic routing, real-time monitoring, and automatic runtime optimization that reduces token costs; none of those core behaviors appear in this chunk. Additionally, the code accesses the local filesystem to persist cache data, which is an undeclared resource use. Because the actual behavior is materially narrower than the declared purpose and includes undeclared disk access, this is a mismatch.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
96% confidence
Finding

The description presents a broad optimization system that actively saves tokens through caching, routing, monitoring, and automatic runtime optimization. The supplied code chunk is much narrower: it monitors recorded cost values, calculates mean/std/CV-based thresholds, detects anomalies, stores historical cost data in a local JSON file, and returns alerts/statistics. While 'real-time monitoring' and cost monitoring are partially aligned, the major advertised behaviors—smart caching for requests, dynamic routing, and automatic optimization to reduce costs—are absent from this code. Additionally, the code uses filesystem read/write persistence, which is not declared in permissions.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The declared description emphasizes token savings through caching, dynamic routing, runtime optimization, and cost monitoring. The supplied code does include a limited cost-monitoring aspect, but its primary behavior is operational governance: controlling whether skills or plans may run, enforcing hard spend limits, tracking anomalies, and persisting these controls on disk. It does not implement caching, model routing, or automatic optimization, which are central to the description. It also introduces undeclared control capabilities and filesystem persistence that are materially different from a simple cost-optimization helper.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
86% confidence
Finding

This markdown file advertises '一键管控 - 禁用技能、设置预算', which implies actions that can affect system behavior or availability, but it provides no user warning about the operational impact of disabling skills or changing budgets. Under SQP-2 for markdown files, descriptions should disclose behaviors that could affect user data, privacy, or system integrity.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill name and user-facing description are written entirely in Chinese, including the tagline and usage framing, with no indication that users may interact in another language. This can violate a language/locale policy when a skill imposes a specific language without explicit user choice or documented regional justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The file's natural-language interface strings, help text, and status messages are presented only in Chinese, with no indication that the tool is region-specific or that users can opt into another language. This can violate language/locale policy requirements when a skill forces a specific language without user choice.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The cache persists arbitrary cached values to disk in plaintext JSON under a predictable local path, with no consent mechanism, minimization, or protection controls. In a token-cost optimization skill, cached content may include prompts, model outputs, API-derived data, or other sensitive user content, so local disclosure risk is materially elevated if the host is shared, backed up, or inspected.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The code invokes a host shell command (ollama list) to inspect locally installed models, which is a privileged environmental probe unrelated to pure cost calculation logic. Even though the command string is static and not obviously user-controlled, it expands the skill's capabilities into local system inspection and creates avoidable privacy and policy risk if this skill runs in sensitive environments.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill silently executes a subprocess to enumerate local Ollama models without any disclosure, consent, or runtime notice. In an agent skill whose stated purpose is token cost optimization, undisclosed host inspection is dangerous because it can reveal local environment details and violate user expectations or sandbox boundaries.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The manifest describes a cost-optimization skill focused on caching, routing, and real-time monitoring for token savings. This file goes beyond monitoring by implementing active operational controls that can disable skills and pause/resume plans, which changes execution behavior rather than merely optimizing or observing cost usage.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
86% confidence
Finding

The natural-language identifier in the file header is written only in Chinese ("Token省钱管家"), which can constitute a language policy issue if the skill is expected to be usable across locales without explicit user opt-in. There is no surrounding comment or indication that this skill is intentionally region-specific or that users can choose their preferred language.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
95% confidence
Finding

The user-facing strings in this installer are exclusively in Chinese, including status and usage messages. This imposes a specific language on users without offering a choice or documenting a justified locale constraint, which matches the language/locale policy violation category.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
88% confidence
Finding

The file's natural-language documentation and inline comments are written in Chinese, with no indication that language choice is optional or user-selected. Under the stated policy, forcing a specific language without opt-in can be a locale/language policy violation.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
97% confidence
Finding

The inline comment says the cache uses an LRU policy, but get() does not refresh entry order and eviction simply removes the first key in Map iteration order. That behavior is FIFO/insertion-order eviction rather than least-recently-used, so the documentation contradicts the actual cache behavior.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The module description and inline documentation are written in Chinese, and the file provides no indication that language choice is configurable or that the skill is intentionally limited to a Chinese-speaking context. This can violate language/locale policy when a skill implicitly enforces one language without user opt-in.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
79% confidence
Finding

The natural-language description "OpenClaw Token省钱管家" is Chinese-only, and all user-facing comments in this file are also Chinese. That suggests a language-specific experience without any visible opt-in, fallback, or documented justification for restricting locale, which may conflict with language-choice policy requirements.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The file's human-readable comments and descriptions are written only in Chinese, which imposes a specific language choice in natural-language content without indicating user choice or locale-specific justification. Under the stated policy, language constraints should be opt-in or clearly justified.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
89% confidence
Finding

This code file contains natural-language comments and user-facing return strings in Chinese, indicating a fixed language choice rather than offering a locale option. The policy explicitly calls for flagging language or locale constraints when the skill forces a specific language without user opt-in.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

Low
Category
Not specified by scanner
Confidence
82% confidence
Finding

The manifest emphasizes intelligent caching, dynamic routing, and real-time cost monitoring, but this code persists control state for disabled skills and paused plans to disk. Long-lived administrative state for workflow/skill control is a broader operational capability than the user-facing description suggests.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
90% confidence
Finding

The natural-language description is entirely in Chinese and gives no indication that the skill is region-specific or that users can choose another language. Under the stated policy, forcing a specific language without opt-in is a locale-policy concern.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

The manifest description includes Chinese text only ("OpenClaw Token省钱管家") with no indication that users can choose their preferred language or locale. This can violate language/locale policy when a skill presents itself in a single language without opt-in or documented regional justification.

Content

No source excerpt is available for this finding.