Back to skill

Security audit

Skill Publisher

Security checks for vulnerabilities and agentic risk

Overview

This skill is purpose-aligned for publishing local skills to ClawHub, but it asks the agent to run unpinned npm CLI commands for login and publishing and gives shell-command templates that can be unsafe with untrusted metadata.

Review this skill before installing. Only use it when you intentionally want to publish a specific local skill to ClawHub, verify the local path and metadata first, avoid secrets in the skill directory, and prefer a pinned or vetted `clawhub` CLI invocation instead of unversioned `npx` commands.

Vulnerability Patterns
  • Insecure DependenciesIntroduces malicious components through unsafe dependency sources
  • 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
Findings (2)

T08 · Insecure Dependencies

Warning
Location
SKILL.md:17
Finding

Unpinned npm Package Is Executed Through npx

Content
View full analysis
Remediation
View remediation
whoami ``` 2. Prefer a locally installed dependency governed by a committed lockfile rather than downloading executable code during each skill invocation. 3. Configure and enforce an approved npm registry instead of trusting arbitrary user or environment registry settings. 4. Verify package provenance and integrity before execution, including package signatures or trusted integrity hashes where supported. 5. Review and update the pinned version through a controlled dependency-update process. 6. Avoid automatically accepting installation prompts unless the package name, version, registry, and integrity have already been validated. ]]>

T09 · Insecure Skill Coding Practices

Error
Location
SKILL.md:38
Finding

Publication Metadata Can Be Unsafely Interpolated into Shell Commands

Content
View full analysis
--slug --name "<显示名称>" --version 1.0.0 --tags "<标签1,标签2>" ``` ```bash cd ~ && npx clawhub publish --slug --version <新版本> --changelog "<更新说明>" ``` In these templates, the localized placeholders represent the display name, tags, new version, and changelog respectively. ### Technical Analysis The templates place several variable values directly into shell command text. The skill path, slug, and version placeholders are unquoted. The display name, tags, and changelog are enclosed in double quotes, but double quotes do not prevent shell command substitution such as `$(command)` or backtick expressions. No instructions require strict input validation, shell escaping, rejection of metacharacters, or execution through an argument-array API. If an implementation substitutes user-controlled values verbatim and executes the resulting string through a shell, those values can alter the command's structure or trigger command substitution. For example, a changelog containing `$(attacker-command)` remains executable even when inserted between double quotes. An unquoted path containing a command separator can terminate the intended command and append another command. ### Attack Path 1. An attacker influences publication metadata, such as the requested skill path, slug, display name, tags, version, or changelog. 2. The attacker supplies shell syntax in that value. Examples include command substitution in a changelog or a command separator in an unquoted path. 3. The agent replaces the documented placeholder with the supplied value. 4. The agent executes the assembled command through a shell. 5. The shell interprets the injected syntax before or while launching `npx`. 6. The attacker's command runs with the privileges and environmen ...[truncated 922 chars]
Remediation
View remediation
", "publish", validatedPath, "--slug", validatedSlug, "--version", validatedVersion, "--changelog", validatedChangelog] shell: false ``` ]]>
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (14)

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The activation condition 需要将本地技能发布到 ClawHub 时触发 is ambiguous and does not define precise invocation boundaries, making accidental activation more likely. In context, ambiguity matters because the skill can lead to login, inspection, and upload actions affecting external state.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The skill description says it uploads local skills to ClawHub but omits a clear warning that local skill contents are transmitted to an external marketplace. This missing disclosure can undermine informed user consent and increase the chance that sensitive files, embedded secrets, or proprietary content are published unintentionally.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
96% confidence
Finding

The skill instructs use of npx clawhub without pinning a package version, so execution depends on whatever package version is currently resolved from the registry at runtime. If the upstream package is compromised, replaced, or unexpectedly changed, the agent could execute attacker-controlled code during login, inspection, or publish operations on the user's machine.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
96% confidence
Finding

Running npx clawhub login without a pinned version introduces a supply-chain execution risk because npx may fetch and run the latest package code from the npm registry. This is especially sensitive in a login flow, where compromised code could steal authentication tokens or redirect the user to a phishing flow.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
96% confidence
Finding

The unpinned npx clawhub inspect <slug> command allows arbitrary upstream package changes to affect local execution. Because this command may be run automatically as part of the publishing workflow, a compromised package could exfiltrate local metadata, credentials, or alter decision-making before publication.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

The publish command executes an unpinned npm package in a workflow that reads local files and uploads them externally, which materially increases the blast radius of a supply-chain compromise. Malicious package code could tamper with the published artifact, steal unpublished skill contents, or run arbitrary code locally.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

This update publish path again relies on npx clawhub without version pinning, exposing a release operation to registry-side compromise or unexpected upstream behavior changes. Since the command processes local skill files and changelog text, a malicious package could exfiltrate repository contents or publish altered material.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The command reference includes npx clawhub whoami without a pinned version, normalizing unsafe dynamic package execution throughout the skill. Even informational commands can become an attack vector if the package is malicious or unexpectedly modified upstream.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
95% confidence
Finding

Referencing npx clawhub login without version pinning is a true vulnerability because it couples an authentication step to untrusted latest-package resolution. In the context of login, compromise could directly expose credentials, session tokens, or mislead the user into authorizing a rogue service.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The inspect reference remains unpinned and therefore inherits the same supply-chain execution risk as the operational examples above. While less sensitive than publish or login, it still executes code fetched based on mutable upstream state.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

The generic publish reference without a pinned version is dangerous because it encourages repeated execution of mutable third-party code in a workflow that reads and uploads local content. In this skill's context, that directly affects confidentiality and integrity of local skill artifacts and release outputs.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The npx clawhub sync reference is also unpinned, creating the same risk of arbitrary code execution from a mutable npm package source. Sync operations may access multiple local skills, so compromise could widen data exposure beyond a single project.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

Listing 发布到市场 in the trigger list broadens matching to common language unrelated to ClawHub, increasing the chance of unintended invocation. Because the skill's purpose is to upload local artifacts to an external marketplace, mistaken activation can have real confidentiality and integrity consequences.

Content

No source excerpt is available for this finding.

Vague Triggers

Low
Category
Not specified by scanner
Confidence
84% confidence
Finding

The activation condition 需要将本地技能发布到 ClawHub 时触发 is ambiguous and does not define precise invocation boundaries, making accidental activation more likely. In context, ambiguity matters because the skill can lead to login, inspection, and upload actions affecting external state.

Content

No source excerpt is available for this finding.