Back to skill

Security audit

Solo Forge

Security checks for vulnerabilities and agentic risk

Overview

SoloForge is a coherent China-market business assistant, but it can persist and automatically change client, project, and payment records without strong confirmation or privacy safeguards.

Review this before installing if you will store real customer, contract, or payment information. Use it in a dedicated workspace, keep data/ out of source control and cloud sync unless intended, confirm every proposed write to clients.json or deals.json, and pin or verify the installer instead of relying on @latest.

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
README.md:102
Finding

Unpinned Package Execution in Installation Instructions

Content
View full analysis

Vulnerability Details

File Location: README.md, lines 102–105
Vulnerability Type: Unpinned third-party package execution
Risk Level: Medium

Vulnerable Code

bash
# ClawHub ecosystem (OpenClaw / QClaw / WorkBuddy)
npx clawhub@latest install Qiu-Difeng/solo-forge

Technical Analysis

The documented installation command instructs users to execute the mutable latest release of the third-party clawhub package through npx. Because neither an exact package version nor an integrity hash is specified, the package executed during installation may differ from the version reviewed when this Skill was audited.

npx can download and execute package code locally. Consequently, compromise of the package publisher, package registry, release pipeline, or a future latest release could introduce arbitrary behavior into the documented installation process.

The repository itself contains no evidence that the current clawhub package is malicious. The issue is the absence of dependency pinning and integrity verification, which creates a supply-chain trust boundary outside the audited project.

Attack Path

  1. An attacker compromises the clawhub package publisher account, registry entry, or release pipeline.
  2. The attacker publishes a malicious release that becomes the package's latest version.
  3. A user follows the installation instructions and runs npx clawhub@latest install Qiu-Difeng/solo-forge.
  4. npx retrieves and executes the attacker-controlled package release.
  5. The malicious package performs actions with the operating-system privileges of the user running the command.

Impact Assessment

Successful exploitation could permit arbitrary code execution under the installing user's account. Depending on that account's permissions and the malicious package behavior, the attacker could access or modify user-readable files, steal local credentials, alter installed Agent skills or configuration, execute network requests, o ...[truncated 246 chars]

Remediation
View remediation

Remediation Suggestions

  1. Replace @latest with an exact, reviewed package version:
    bash
    npx clawhub@X.Y.Z install Qiu-Difeng/solo-forge
    
  2. Document the expected package publisher and official registry source so users can identify namespace substitution or registry redirection.
  3. Publish and verify a cryptographic integrity value or signed release artifact before execution.
  4. Use a lockfile or equivalent immutable dependency record where the installation environment supports one.
  5. Establish a controlled process for reviewing and deliberately updating the pinned version.
  6. Advise users not to run the installer with administrator or root privileges unless strictly required.
  7. Where practical, offer a non-executing installation method based on a pinned commit or signed release archive.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (23)

Vague Triggers

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

Using 帮我 as a trigger phrase is dangerously broad because it overlaps with ordinary conversation and can cause the skill to activate unintentionally in unrelated contexts. In an agent ecosystem with routing or auto-invocation, this can lead to unintended access to business data, writes to local JSON state, or generation of legal/commercial documents without clear user intent.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The README states the skill is aimed at the China market, with contract terms based on Chinese law and pricing in RMB, and later emphasizes Chinese business-culture adaptation. This imposes a locale-specific operating context without offering a user choice or explicit opt-in path for other languages/locales.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The README emphasizes persistent local storage of client, deal, and contract-related data but does not clearly warn users that these files may contain sensitive commercial and personal information. This is risky because users may unknowingly store confidential business details in plaintext local JSON, increasing exposure through backups, shared machines, source control mistakes, or other local compromise.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
70% confidence
Finding

npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill explicitly instructs direct creation and modification of files under data/ and even says to create missing files, but it does not require explicit user confirmation before mutating state. In an agent context, that can lead to unintended persistence or silent alteration of business records from ambiguous prompts, which is risky because these files contain client and deal data used in later decisions and documents.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The trigger list includes vague phrases such as "算一下", "多少钱", "怎么回", "跟一下", and "做个计划", which commonly appear in ordinary conversation outside this skill's intended scope. Because the file presents these as routing triggers without clear exclusion conditions or negative examples, the skill may activate the wrong module for routine user messages.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The instruction mandates a specific locale convention: all monetary output must use RMB formatting and the surrounding skill content is written as if Chinese-language operation is required. This is a natural-language locale policy constraint, but the file does not provide user opt-in or state that the skill is intentionally limited to a China-specific business context.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

This JSON file contains user-facing natural-language content exclusively in Chinese, including company descriptors, notes, and research fields. Under the policy, forcing a specific language without user opt-in is a natural-language locale violation unless the regional constraint is documented and justified, which is not present here.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The config sets the currency to "人民币", which implies a fixed Chinese locale/currency context. For a general-purpose skill configuration, this is a natural-language locale policy constraint without any indication of user choice, opt-in, or region-specific justification.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The module explicitly directs the agent to read and write customer records and research data to a local JSON file, but it gives no instruction to notify the user that potentially sensitive business contact information will be persisted. This creates a real privacy and data-governance risk because users may provide names, contacts, notes, and research content assuming they are ephemeral, while the skill silently stores them on disk.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The research workflow instructs the agent to perform external searches using company names and role keywords, but it does not warn the user that these queries may be sent to third-party network services. That is a genuine data-exposure issue because confidential prospecting targets, client interests, or deal-stage context can be inferred from outbound searches without the user's informed consent.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The entire template and generation instruction are written to produce Chinese legal documents, and there is no indication that users may choose another language or locale for output. This is a natural-language policy concern because the skill effectively mandates a specific language without explicit user opt-in.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The entire skill file is written as Chinese-only operational guidance for generating contracts, NDAs, and SOWs, and it does not indicate that users may choose another language or locale. Under the policy, forcing a specific language without opt-in is a natural-language policy violation unless the locale restriction is explicitly justified.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The trigger condition '生成进度报告、验收报告、满意度调查时加载此模块' is broad and overlaps with ordinary business-language requests, so the module may activate unintentionally for loosely related prompts. In this skill, accidental activation can expose deal-specific payment logic and steer outputs toward催款/验收 strategies, causing inappropriate document generation or unintended use of project data.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The activation condition uses a very common user phrase pattern ("客户说XXX,怎么回") that can appear in many ordinary conversations, making this module likely to trigger outside its intended narrow context. Over-broad triggering can cause the agent to inject negotiation playbooks when the user did not explicitly request this skill behavior, leading to unintended guidance, context hijacking, or incorrect responses in adjacent workflows.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The file is entirely written as prescriptive Chinese-only templates and instructions, with no indication that language should follow user preference or that this is a region-specific skill. Under the policy criteria, forcing a specific language without user opt-in is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

This markdown file presents all skill content and instructions exclusively in Chinese, including headings, templates, and behavioral guidance. Under the language/locale policy rule, forcing a specific language without user opt-in or an explicit region-specific justification is a natural-language policy violation.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
89% confidence
Finding

The module states that all project data is read and written directly to JSON files, but it does not describe any user warning, preview, or safeguards around destructive or persistent modifications. In this skill's context, those files hold operational deal and client data, so silent writes can create hard-to-detect corruption, overwrite mistakes, or unauthorized changes to business records.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
92% confidence
Finding

The operation mapping uses broad natural-language triggers such as '添加项目/新建项目' and '更新项目/已收款X元' without requiring confirmation, disambiguation, or structured arguments. In a document-generating business assistant that directly writes JSON project records, ambiguous phrasing can cause unintended state or financial data changes, leading to integrity issues and downstream mistakes in invoices, reminders, and reports.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The status transition rules treat casual phrases like '做完了', '验收通过了', and '项目取消了' as automatic triggers for changing project lifecycle state and updating timestamps. Because these states drive collection actions, settlement logic, and completion tracking, accidental or context-misread transitions can materially affect contractual workflow and financial follow-up.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
94% confidence
Finding

The document specifies that certain user phrases automatically update status and updated_at, but it does not require confirmation or warn about the consequences of those changes. Since status changes can trigger acceptance, payment collection, cancellation handling, and legal-style reminder escalation, automatic updates increase the risk of unintended business actions based on misinterpreted conversation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
83% confidence
Finding

This JSON file contains user-facing text values entirely in Chinese, including client/project labels and milestone names, but provides no indication that the locale is optional or intentionally restricted to a Chinese-language context. Under the policy rules, hard-coded language constraints without opt-in or documented justification can constitute a natural-language policy issue.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
87% confidence
Finding

This markdown file contains user-facing natural-language templates entirely in Chinese, and there is no indication that the skill is region-specific or that users can opt into another language. Under the policy, forcing a specific language without user opt-in is a natural-language policy concern.

Content

No source excerpt is available for this finding.