Back to skill

Security audit

skillhub-auto-installer

Security checks for vulnerabilities and agentic risk

Overview

The skill presents itself as search-only, but the package also contains scripts that can install remote skills and inspect local skill directories.

Review this skill carefully before installing. Its search function is simple, but the package includes under-disclosed install and local-audit scripts that can copy third-party skills into the active agent environment using unpinned npm tooling. Treat it as an installer/security-checking utility, not a read-only search skill, and avoid running the installer scripts unless you have independently reviewed the target skill and trust the npx package source.

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 (3)

other

Warning
Location
scripts/install.sh:32
Finding

Installation Capability Contradicts the Declared Search-Only Scope

Content
View full analysis
Remediation
View remediation

T08 · Insecure Dependencies

Error
Location
scripts/search.sh:17
Finding

Unpinned Runtime Dependency Execution Through npx

Content
View full analysis
/dev/null || { echo "搜索失败,请检查网络连接" exit 1 } ``` The installation scripts use the same unsafe dependency resolution pattern: ```bash SKILLS_API_URL=https://skills.volces.com/v1 npx -y skills add "$URL" -s "$SKILL_NAME" -a openclaw -y --copy ``` ### Technical Analysis The scripts invoke the npm package named `skills` through `npx -y` without specifying an exact version. The project contains no lockfile, integrity hash, vendored audited dependency, or other mechanism that binds execution to the version reviewed during this audit. As a result, the effective code executed by a future invocation may differ from the code available when the Skill was reviewed. The `-y` option automatically accepts package installation and suppresses the normal confirmation prompt. This affects even the advertised read-only search flow. Although the script calls the package with a `find` operation, npm package lifecycle behavior and package entry-point code execute locally and are not inherently constrained to read-only behavior. ### Attack Path 1. An attacker compromises the upstream `skills` npm package, its maintainer account, publication credentials, or a dependency used by that package. 2. A malicious or otherwise unsafe release becomes the version resolved by `npx`. 3. A user performs an ordinary Skill search or installation. 4. `npx -y skills` retrieves and executes the currently resolved package without asking for confirmation. 5. Malici ...[truncated 804 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
scripts/secure-install.sh:38
Finding

Secure Installer Proceeds When Pre-Installation Auditing Is Missing or Fails

Content
View full analysis
/tmp/pre-install-audit.json 2>/dev/null || true if [ -f /tmp/pre-install-audit.json ]; then RISK_COUNT=$(cat /tmp/pre-install-audit.json | grep -o '"risk"' | wc -l || echo "0") if [ "$RISK_COUNT" -gt 0 ]; then echo " ⚠️ 发现 $RISK_COUNT 个潜在风险,建议检查" else echo " ✅ 当前环境安全状态良好" fi fi fi echo "" fi # 步骤 2: 安装 Skill echo "⚡ 步骤 2/3: 安装 Skill..." echo "" URL="https://skills.volces.com/skills/$OWNER/$REPO" cd /home/gem/workspace/agent SKILLS_API_URL=https://skills.volces.com/v1 npx -y skills add "$URL" -s "$SKILL_NAME" -a openclaw -y --copy ``` ### Technical Analysis The script is named and presented as a secure installer, but its validation is fail-open: - If SkillSentry is absent, the script explicitly skips the scan and continues. - Audit failures are suppressed with `2>/dev/null || true`. - Detected risks produce only a warning and do not block installation. - The pre-install audit scans the current installed environment, not the remote target that is about to be installed. - Target-specific inspection occurs only after the remote Skill has already been copied into the active Skill directory. The separate `pre-install-check.sh` has a similar reliability weakness at lines 101-106: ...[truncated 1575 chars]
Remediation
View remediation
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Tool MisuseTool Parameter Abuse, Chaining Abuse, Unsafe Defaults
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (53)

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill repeatedly claims it only performs search and does not install or modify local state, yet it includes explicit installation instructions using npx ... skills add ... --copy. This mismatch is dangerous because agents or users may trust the safety claims and overlook that the documented workflow can lead to local changes and execution of untrusted marketplace content.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill repeatedly claims it only performs search and does not install or modify local state, yet it includes explicit installation instructions using npx ... skills add ... --copy. This mismatch is dangerous because agents or users may trust the safety claims and overlook that the documented workflow can lead to local changes and execution of untrusted marketplace content.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill repeatedly claims it only performs search and does not install or modify local state, yet it includes explicit installation instructions using npx ... skills add ... --copy. This mismatch is dangerous because agents or users may trust the safety claims and overlook that the documented workflow can lead to local changes and execution of untrusted marketplace content.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill repeatedly claims it only performs search and does not install or modify local state, yet it includes explicit installation instructions using npx ... skills add ... --copy. This mismatch is dangerous because agents or users may trust the safety claims and overlook that the documented workflow can lead to local changes and execution of untrusted marketplace content.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The manifest states this skill only searches/discovers skills and does not perform installation or local file operations, but this script clearly installs a remote skill and uses --copy to write into the local workspace. That mismatch is security-relevant because users and higher-level tooling may grant trust or permissions based on the declared limited behavior, while the actual behavior performs environment changes and may introduce unreviewed code locally.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

This script changes into a fixed workspace and installs/copies content from a remote source into the local environment, which is unnecessary for a skill advertised as search-only. In this context, unjustified write capability is especially dangerous because it expands the attack surface from passive discovery to active supply-chain and filesystem modification behavior.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
96% confidence
Finding

The skill metadata says it only provides search and performs no installation or local file operations, but this script inspects the local environment, probes network reachability, searches local filesystem paths, and executes a local audit script. That mismatch is dangerous because users and policy engines may grant trust based on the declared scope while the actual behavior expands into local system interaction.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
94% confidence
Finding

A search-only skill should not contain pre-install security-auditing logic that traverses local paths and runs another script, especially when that script is discovered dynamically from multiple relative locations. This broadens the trust boundary and can result in unexpected execution of local code, including a malicious audit.sh planted in one of the searched directories.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The manifest explicitly claims no local file operations, but the script reads from and writes to /tmp and the local skills directory during auditing and reporting. This discrepancy weakens trust boundaries and can mislead users, orchestration systems, or policy engines into allowing a file-writing skill under search-only assumptions.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

The skill metadata says it only searches/discovers skills and does not perform installation or local file operations, but this script actually installs a skill into the local environment. That mismatch is security-relevant because users and higher-level agents may grant it search-only trust while it performs side-effectful actions, enabling unexpected code import and persistence.

Content

No source excerpt is available for this finding.

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

Including installation capability is unjustified for a skill advertised purely as a discovery/search tool, and it materially increases risk by downloading and installing remote content. In context, this is more dangerous because the benign-seeming marketplace-search description can cause users or agents to underestimate the privilege and side effects of invoking it.

Content

No source excerpt is available for this finding.

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
100% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · scripts/secure-install.sh (reported line 107)May include surrounding context.

sh
echo "   📁 发现 $SCRIPT_COUNT 个脚本文件"
            
            # 检查脚本中是否有可疑命令
            SUSPICIOUS=$(grep -r "rm -rf /\|curl.*|.*sh\|wget.*|.*sh\|bash -c.*\$(curl" "$SKILLS_DIR/$SKILL_NAME/scripts/" 2>/dev/null || true)
            if [ ! -z "$SUSPICIOUS" ]; then
                echo "   🚨 警告: 发现可疑命令模式,请人工审查!"
                echo "$SUSPICIOUS" | head -3

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
100% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · scripts/security-check.sh (reported line 120)May include surrounding context.

sh
echo "   📁 发现 $SCRIPT_COUNT 个脚本文件"
            
            # 检查脚本中是否有可疑命令
            SUSPICIOUS=$(grep -r "rm -rf /\|curl.*|.*sh\|wget.*|.*sh\|bash -c.*\$(curl" "$SKILLS_DIR/$SKILL_NAME/scripts/" 2>/dev/null || true)
            if [ ! -z "$SUSPICIOUS" ]; then
                echo "   🚨 警告: 发现可疑命令模式,请人工审查!"
                echo "$SUSPICIOUS" | head -3

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
95% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · scripts/secure-install.sh (reported line 107)May include surrounding context.

sh
echo "   📁 发现 $SCRIPT_COUNT 个脚本文件"
            
            # 检查脚本中是否有可疑命令
            SUSPICIOUS=$(grep -r "rm -rf /\|curl.*|.*sh\|wget.*|.*sh\|bash -c.*\$(curl" "$SKILLS_DIR/$SKILL_NAME/scripts/" 2>/dev/null || true)
            if [ ! -z "$SUSPICIOUS" ]; then
                echo "   🚨 警告: 发现可疑命令模式,请人工审查!"
                echo "$SUSPICIOUS" | head -3

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
100% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · scripts/secure-install.sh (reported line 107)May include surrounding context.

sh
echo "   📁 发现 $SCRIPT_COUNT 个脚本文件"
            
            # 检查脚本中是否有可疑命令
            SUSPICIOUS=$(grep -r "rm -rf /\|curl.*|.*sh\|wget.*|.*sh\|bash -c.*\$(curl" "$SKILLS_DIR/$SKILL_NAME/scripts/" 2>/dev/null || true)
            if [ ! -z "$SUSPICIOUS" ]; then
                echo "   🚨 警告: 发现可疑命令模式,请人工审查!"
                echo "$SUSPICIOUS" | head -3

Intent-Code Divergence

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The script's own header states it is a security-check tool for installed skills, directly contradicting the manifest's claim that the skill only searches the marketplace and performs no local operations. This discrepancy is a strong indicator of undeclared behavior and increases the risk of users granting trust under false assumptions.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
99% confidence
Finding

This code enumerates installed skills, reads local files such as SKILL.md, and traverses local script directories despite the skill being described as search-only with no local file operations. Hidden local inspection broadens access to potentially sensitive filesystem metadata and violates least surprise.

Content

No source excerpt is available for this finding.

Tool Parameter Abuse

High
Category
Tool Misuse
Confidence
90% confidence
Finding

Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).

Content

Scanner excerpt · scripts/security-check.sh (reported line 120)May include surrounding context.

sh
echo ""
            echo "🔍 安全检查:"
            
            DANGEROUS=$(grep -rE "(rm -rf /|rm -rf \*|> /etc/|chmod 777 /|curl.*\|.*sh|wget.*\|.*sh)" "$TARGET_SKILL_DIR/scripts/" 2>/dev/null | grep -v "^#" || true)
            if [ ! -z "$DANGEROUS" ]; then
                echo "   🚨 发现高危命令模式:"
                echo "$DANGEROUS" | head -5 | sed 's/^/      /'

Context-Inappropriate Capability

High
Category
Not specified by scanner
Confidence
97% confidence
Finding

Executing a separate local 'audit.sh' from another skill/tool is unjustified for a marketplace discovery skill and creates a second-stage execution path. If that tool is malicious, outdated, or replaced, this script becomes a launcher for arbitrary local code under the user's trust boundary.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
Not specified by scanner
Confidence
98% confidence
Finding

The manifest says this skill only performs search and does not execute installs or local file operations, but this script runs a local audit tool. That mismatch is dangerous because it expands the skill's real capability beyond declared scope, reducing user ability to make informed trust decisions.

Content

No source excerpt is available for this finding.

Undeclared Tool Scope

Medium
Category
MCP Least Privilege
Confidence
83% confidence
Finding

The skill references shell-based execution paths via npx commands in the documented workflow, but it does not declare any explicit tool scope or permissions. That creates an authorization ambiguity where an agent or reviewer may assume the skill is read-only while it actually depends on command execution, increasing the chance of unintended command use.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
83% confidence
Finding

The manifest description and keywords are presented exclusively in Chinese, and the rest of the file continues in Chinese for the substantive instructions. Under the policy rule, forcing a specific language without user opt-in or documented justification is a natural-language policy concern.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

Using npx -y skills without pinning a specific package version causes execution of whatever version is currently resolved from the registry. This introduces supply-chain risk: a compromised or malicious update could change behavior and execute attacker-controlled code on the host.

Content

No source excerpt is available for this finding.

Rp1

Medium
Category
MCP Rug Pull
Confidence
97% confidence
Finding

The install example uses npx -y skills add without a pinned version, which is especially dangerous because it combines remote package resolution with a command that modifies the local environment. An attacker controlling the package supply chain or a malicious update could gain code execution during installation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
84% confidence
Finding

This markdown file is entirely written as normative user-facing guidance in Chinese, but it does not state that the skill is China-region-specific or offer an opt-in language/locale choice. Under the language/locale policy rule, a skill that effectively forces one language for use or understanding can be a natural-language policy violation unless the constraint is documented and justified.

Content

No source excerpt is available for this finding.