Back to skill

Security audit

Anti-Laziness Protocol

Security checks for vulnerabilities and agentic risk

Overview

This is a Chinese-language quality-control skill that is broadly scoped and asks for local progress notes, but I found no hidden code, network behavior, credential handling, or destructive actions.

Install this if you want a stricter Chinese-language workflow for code review, research, writing, and validation. Be aware it may activate broadly and may cause the agent to create local progress notes containing findings from your work; avoid using it on sensitive projects unless you control where those notes are stored and when they are deleted.

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

Note
Location
SKILL.md:66
Finding

Unsafe Handling of Temporary and Progress Files

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:66-69
Vulnerability Type: Unspecified temporary-file storage and lifecycle
Risk Level: Low

Vulnerable Instruction

markdown
### 7. Process Annotations and Memory Consolidation

- During long tasks, every key finding must be written to a temporary file.
- After each step, record a one-sentence summary in a progress file.

The snippet is an English translation of the instructions at the specified location.

Technical Analysis

The Skill requires the Agent to write findings to temporary and progress files but does not specify a secure directory, restrictive file permissions, safe filename generation, content-minimization rules, retention limits, or cleanup procedures.

These files can contain sensitive information gathered during source-code reviews, including vulnerability details, proprietary implementation information, local paths, configuration data, and excerpts derived from private project material. If the Agent uses a shared temporary directory, predictable filename, permissive default mode, or persistent workspace location, another local process or user may be able to read the resulting artifacts.

The project does not contain code that deliberately exfiltrates these files, creates system persistence, or elevates privileges. Exploitation therefore depends on the runtime environment and the Agent's implementation of the instruction.

Attack Path

  1. A user asks the Agent to analyze a private repository or other sensitive material.
  2. Following SKILL.md:66-69, the Agent records each key finding and progress summary in unspecified files.
  3. The Agent creates those files in a shared or persistent location using default or overly permissive access controls.
  4. Sensitive findings remain available after the task or become accessible to another local user or process.
  5. The local principal reads the residual files and obtains information deri ...[truncated 579 chars]
Remediation
View remediation

Remediation Suggestions

  1. Make progress-file creation optional and require explicit user approval before writing analysis artifacts.
  2. Store files only in a task-scoped private directory outside shared temporary locations.
  3. Create directories and files with owner-only permissions, such as directory mode 0700 and file mode 0600.
  4. Generate unpredictable filenames using secure temporary-file APIs rather than fixed or user-controlled names.
  5. Record only minimal progress metadata; do not copy secrets, credentials, source contents, personal data, or complete vulnerability evidence into temporary files.
  6. Define a strict retention period and delete all temporary artifacts when the task ends, including during error handling or cancellation.
  7. Clearly distinguish temporary task notes from persistent Agent memory and prohibit cross-session retention by default.
  8. If durable records are required, disclose their exact location, contents, access controls, and retention policy to the user before creation.
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 (6)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest description is written to require use of the Chinese-language protocol and the document consistently mandates Chinese-specific labels such as "标注 [待验证]" and localized phrasing, without offering a user language choice. This can violate language/locale policy when the user has not opted into Chinese output or terminology.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The trigger conditions are very broad: phrases like '任何需要验证的工作' and '用户要求"详细分析"' can match a large portion of normal user requests, causing the skill to activate outside its intended scope. Over-broad auto-invocation can override normal task handling, increase prompt surface area, and impose hidden behavioral constraints such as mandatory workflow and self-check steps that the user did not request.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

该规范在“必须的风格”中直接要求“口语化”,并要求删除“学术腔”,属于对输出语言/文体的强制性约束。文件未说明这是仅适用于特定场景的可选规范,也未提供用户选择或例外条件,因而构成自然语言层面的语言/风格政策限制。

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The file is entirely written as a mandatory normative specification in Chinese, beginning with the title at L001, and nowhere indicates that language is optional or that users may choose another language. Under the policy rule, forcing a specific language without opt-in or documented justification is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
91% confidence
Finding

The manifest description is entirely in Chinese, which implies the skill is presented in a single language without indicating any user language choice or opt-in. Under the policy, language constraints should either offer a choice or be clearly documented as region-specific and 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 markdown file contains natural-language instructions exclusively in Chinese and does not indicate that users may choose another language. Under the policy rule for language or locale constraints, forcing a specific language without user opt-in is a reportable issue.

Content

No source excerpt is available for this finding.