Back to skill

Security audit

teamarchitect-skill

Security checks for vulnerabilities and agentic risk

Overview

This skill is not malicious, but it can install or create other skills in the local agent environment using broad marketplace/GitHub recommendations that are not pinned or provenance-checked.

Review each recommended skill before installation. Do not approve batch installs unless the exact source, publisher, version or commit, permissions, and expected file changes are shown. Prefer pinned, trusted packages and avoid installing newly discovered GitHub or marketplace skills by name alone.

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

Error
Location
SKILL.md:90
Finding

Unverified and Unpinned Third-Party Skill Installation

Content
View full analysis

Vulnerability Details

File Location: SKILL.md:90-103, with supporting evidence in _meta.json:10-14, README.md:10, and skills-index.md:7-104
Vulnerability Type: Third-party supply-chain risk caused by unverified skill discovery and installation
Risk Level: High

Evidence

The following is a faithful English rendering of the installation policy defined in SKILL.md:

text
## Skill Installation Strategy

For each recommended skill:
1. Already installed locally → Skip
2. Available on ClawdHub → Recommend installation
3. Available as open source on GitHub → Recommend downloading
4. Not available from either source → Ask whether to create a new skill

## Installation Execution Stage
- Core skills: Batch install or create after user confirmation
- Optional skills: Install according to the user's selection
- Provide alternatives when installation fails
- Missing skills: Ask whether to create them immediately

The project metadata grants the capabilities needed to access and install remote content:

json
"permissions": {
  "bash": true,
  "network": true,
  "filesystem": "skills-directory"
}

The indexed dependencies are identified primarily by generic names and broad source labels such as ClawdHub rather than by immutable package identities. Examples include data-analyzer, web-scraper, code-reviewer, security-scanner, excel-handler, and test-runner.

Technical Analysis

The skill instructs an agent to discover and install third-party skills from ClawdHub or GitHub. However, it does not require:

  • An allowlist of trusted publishers or repositories
  • Exact repository URLs or verified package identifiers
  • Immutable commit hashes or locked versions
  • Cryptographic signatures or checksums
  • Source-code review before installation
  • Manifest and permission review
  • Detection of typosquatting or dependency confusion
  • Separate approval for each r ...[truncated 2611 chars]
Remediation
View remediation

Remediation Suggestions

  1. Replace generic marketplace names with exact package coordinates, including the verified publisher, canonical repository URL, and immutable commit hash or version.
  2. Maintain an explicit allowlist of trusted registries, organizations, repositories, and signing identities.
  3. Require cryptographic signature verification or a trusted checksum before installing any downloaded artifact.
  4. Review the complete source, manifest, installation hooks, scripts, and requested permissions before approval.
  5. Display provenance information for every recommendation, including publisher, repository, revision, checksum, permissions, and audit status.
  6. Require separate approval for each resolved dependency rather than one approval for an entire batch.
  7. Run newly installed skills in a sandbox with no credentials and minimal network, command, and filesystem access.
  8. Apply least privilege by enabling Bash, network, and filesystem permissions only during the specific installation operation that requires them.
  9. Detect package-name similarity, ownership changes, unexpected source changes, and typosquatting before installation.
  10. Generate and retain a lockfile or installation record containing immutable revisions and integrity hashes for reproducible audits and upgrades.
  11. Revise the documentation's “safe installation” claim unless these provenance, integrity, and permission controls are enforced.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (8)

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The README explicitly advertises '核心技能自动安装' (core skills auto-install) but does not mention confirmation prompts, scope of changes, permissions, or rollback behavior. In a skill that recommends and installs additional capabilities, silent or assumed installation can lead to unexpected system modifications or expansion of the agent’s privileges without informed user consent.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The usage section says the skill will '协助安装和配置' but omits any safety, privacy, or approval guidance. Because this skill is positioned as an orchestrator for multi-agent teams and skill recommendations, users may be led to approve installations or configuration changes without understanding data exposure, permission implications, or the trustworthiness of recommended skills.

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 guides users through installing, downloading, or creating skills from local storage, a marketplace, GitHub, and ad-hoc creation, but does not require any trust verification, provenance checks, permission disclosure, or warning that these actions modify the local environment. This is dangerous because it normalizes pulling executable capability packages from external sources and writing them into the agent environment, which can enable supply-chain compromise or unsafe local changes if a recommended skill is malicious or tampered with.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
88% confidence
Finding

The manifest description is broad and conversationally oriented, describing a skill that learns user requirements, designs multi-agent architectures, and recommends or installs skills without clear trigger constraints. In combination with the declared bash, network, and skills-directory filesystem permissions, this can cause the skill to be invoked for generic planning or advisory requests and then perform higher-risk actions than the user explicitly intended.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
93% confidence
Finding

The description is written entirely in Chinese and presents the skill identity and behavior in that language, with no indication that users may choose another language or that the skill is intended only for a Chinese-speaking context. This can violate language or locale policy when a skill appears to impose a language without explicit opt-in or documented regional justification.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

This markdown file is written entirely in Chinese, including headings, labels, and usage guidance, with no indication that language selection is optional or that the skill index is intended only for a Chinese-speaking context. Under the policy rule, forcing a specific language without user opt-in is a natural-language locale violation.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
73% confidence
Finding

The natural-language instructions and usage guidance are presented entirely in Chinese, and the README does not indicate that the skill supports multiple languages or that Chinese is a justified locale-specific requirement. This can conflict with language-choice policies when users are not given an explicit opt-in or alternative.

Content

No source excerpt is available for this finding.

Natural-Language Policy Violations

Low
Category
Not specified by scanner
Confidence
78% confidence
Finding

The skill title, instructions, prompts, and examples are all presented in Chinese, and there is no indication that users may choose another language or that the skill is intentionally limited to a Chinese-speaking context. Under the policy, forcing a specific language without opt-in is a natural-language policy concern.

Content

No source excerpt is available for this finding.