Back to skill

Security audit

Chatmosp File Organizer

Security checks for vulnerabilities and agentic risk

Overview

The skill is a disclosed ChatMOSP file organizer with scoped task-directory writes; it has documentation inconsistencies and unpinned companion-skill guidance but no evidence of hidden execution, exfiltration, or persistence.

Before installing, be comfortable with the skill creating and copying task files under mosp-for-chatMOSP/OUTPUT/. Verify destination paths before allowing writes, avoid overwriting existing task directories unless intended, and install companion skills only from trusted, verified, preferably pinned ClawHub sources rather than an unspecified GitHub fallback.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:244
Finding

Unpinned External Companion Skills from Insufficiently Specified Sources

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:244-259 and the equivalent companion-skill instructions in SKILL_cn.md:244-259
Vulnerability Type: Unpinned third-party Skill dependencies
Risk Level: Medium

Vulnerable Code Snippet

markdown
## ChatMOSP Companion Skills

This skill is part of the ChatMOSP skill set. For complete functionality, install the companion ChatMOSP skills from ClawHub when they are missing:

- chatmosp-input-coordinator
- chatmosp-parameter-builder
- chatmosp-msr-generator
- chatmosp-kmc-simulator
- chatmosp-literature-search
- chatmosp-file-organizer

If a required companion skill is unavailable, ask the user to install the missing ChatMOSP skill from ClawHub before continuing. Use the GitHub repository only as a fallback.

Technical Analysis

The Skill directs users to install several external companion Skills without specifying immutable versions, verified publisher identities, exact repository URLs, commit hashes, signatures, or integrity checksums. The GitHub fallback is also not tied to a specific organization, repository, release, or commit.

Skill dependencies can contain executable scripts or authoritative instructions that are loaded into an Agent context. Resolving dependencies solely by mutable names creates a supply-chain trust gap: a compromised publisher account, replaced package, malicious namesake, dependency-confusion entry, or modified repository could supply content different from what was originally reviewed.

The project itself contains only documentation files, and the reviewed text does not automatically download or execute a payload. Exploitation therefore depends on a user or orchestration system following the installation recommendation. Nevertheless, the absence of source and version pinning leaves no reliable mechanism for ensuring that the installed dependency is the intended reviewed artifact.

Attack Path

  1. One of the named Cha ...[truncated 1420 chars]
Remediation
View remediation

Remediation Suggestions

  1. Pin every companion Skill to an immutable, reviewed version or content-addressed release.
  2. Identify the expected publisher and provide exact official ClawHub package URLs rather than dependency names alone.
  3. Replace the generic GitHub fallback with exact organization and repository URLs pinned to verified commit hashes.
  4. Publish cryptographic checksums or signed manifests for every approved dependency artifact.
  5. Verify package signatures, publisher identity, version, and checksum before installation or loading.
  6. Maintain an allowlist of approved dependency identifiers and reject namesakes or packages from alternate publishers.
  7. Require explicit user approval after displaying the resolved source, publisher, version, requested capabilities, and integrity result.
  8. Review companion Skill instructions and scripts before granting filesystem, shell, network, credential, or persistent-storage access.
  9. Run third-party Skills with least privilege and isolate them from unrelated files, credentials, tools, and network destinations.
  10. Record dependency provenance and verified hashes in a lockfile or equivalent manifest to make installations reproducible.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (10)

Ae1

High
Category
analysis-evasion
Confidence
100% confidence
Finding

Referenced artifact was not completely inspected

Content

Scanner excerpt · SKILL.md (reported line 15)May include surrounding context.

md
> - If the user writes in Chinese, read `SKILL.md` in this same directory and respond in Chinese.

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · SKILL.md (reported line 1)May include surrounding context.

md
---
name: chatmosp-file-organizer
description: |
  chatMOSP 系统的文件系统管理器。负责 MSR / KMC 任务的目录结构创建、智能命名(按

Hidden Instructions

High
Category
Prompt Injection
Confidence
60% confidence
Finding

Hidden instructions were detected in comments or invisible text. These could contain malicious directives. Manual review is recommended.

Content

Scanner excerpt · SKILL_cn.md (reported line 1)May include surrounding context.

md
---
name: chatmosp-file-organizer
description: |
  chatMOSP 系统的文件系统管理器。负责 MSR / KMC 任务的目录结构创建、智能命名(按

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The trigger language is broad ('after parameter-builder confirms parameters' and similar flow-based conditions) without explicit deny conditions, user-consent requirements, or strict invocation boundaries. For a filesystem-writing skill, ambiguous activation can cause unintended directory creation or file operations in response to loosely related workflow states.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill describes directory creation and secure file operations but does not clearly warn that it will create or modify files on disk. In agent settings, missing disclosure around filesystem side effects increases the risk of silent or surprising writes, especially when combined with broad triggers and cross-skill handoffs.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The skill states that KMC directories must be created under the corresponding MSR directory, but later documents a direct-KMC mode that places a KMC task directly under OUTPUT/. This inconsistency can cause the agent or downstream tools to create files in the wrong location, weakening path-policy assumptions and increasing the chance of unsafe or unintended filesystem operations.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
98% confidence
Finding

The language-routing block is internally contradictory: in a file named SKILL_cn.md, it instructs English users to read SKILL_cn.md while Chinese users should continue using SKILL.md, and the surrounding text/documentation also mismatches which file is authoritative. In an agent skill, conflicting instruction-selection logic can cause the model to load or prioritize the wrong prompt file, leading to incorrect behavior, bypass of expected safeguards, or failure to apply the intended language-specific constraints.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The language-routing instructions are not just editorially inconsistent; they can actively steer the agent to the wrong language-specific instruction source. In prompt-driven systems, ambiguous dispatch logic is dangerous because it can cause the model to ignore the correct operational guidance, producing unreliable or unsafe file-handling behavior.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The document states that KMC directories must be created under the corresponding MSR task directory, but later defines a 'direct KMC' mode that places KMC tasks directly under mosp-for-chatMOSP/OUTPUT/{kmc_task_name}/. This inconsistency can cause an agent to create files in unintended locations, weakening path policy assumptions and making downstream tools operate on the wrong dataset or trust boundary.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
99% confidence
Finding

The file structure block labels SKILL.md as 'Chinese reference' and SKILL_cn.md as 'This file (English)', while earlier language-routing instructions say SKILL.md should be used for English and SKILL.md in the same directory should be read for Chinese, implying inconsistent file-language intent. This is a direct documentation contradiction that can misroute users or other agents.

Content

No source excerpt is available for this finding.