Back to skill

Security audit

Skill Audit & Publish

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because it reports an approval gap in a publishing helper and concerns about network permissions; confidence is medium, and the scan was partial.

Consider the approval gap and undeclared network permission before installing, especially if you may invoke publishing features. The scan covered 7 of 9 components, with 2 partially inspected; obfuscated instruction text and some references could not be fully evaluated, so the evidence is incomplete.

Vulnerability Patterns
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (1)

T09 · Insecure Skill Coding Practices

Warning
Location
scripts/sync_skill_to_github.js
Finding

Directly callable publishing helper does not enforce approval

Content
View full analysis

Impact

Direct invocation can publish local content or replace existing repository files before the user approves the exact changes. Public repository publication can disclose content that has not completed the required review.

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Findings (9)

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · SKILL.md (reported line 14)May include surrounding context.

md
- "network: api.github.com — used only by the bundled sync helper when explicitly invoked"
      - "credentials: GITHUB_TOKEN / GITHUB_PAT environment variables — read at runtime, never stored or logged"
      - "network: api.skillhub.cn — used only by the stage 5b SkillHub upload when the user approves publishing"
      - "credentials: the local SkillHub credential file (~/.skillhub/credentials.json, field user.token) read only during stage 5b to authenticate the upload — held in memory for that single request, never embedded in the skill, logged, or shipped in any package"
      - network
    tags:
      - skill-publishing

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · SKILL.md (reported line 139)May include surrounding context.

md
Two things to know before you consider skipping it:

- **There is no read API.** Every `GET` under `/api/v1/community/skills/*` returns 405. You cannot verify remotely whether a skill is already on SkillHub; use `clawhub inspect <slug> --versions` plus the GitHub mirror as side evidence.
- **Auth is a local credential file, read in memory only.** The upload authenticates with a Bearer token read from the SkillHub credential file (`~/.skillhub/credentials.json`, field `user.token`). The token is held in memory for that single request — never embedded in the skill, logged, or included in any published package.
- **Skipping it is the most common way the unified-version rule breaks.** The skill lands on two platforms, the version registry drifts apart, and the next publish has to guess which number is authoritative.

Load **`references/skillhub-publish.md`** during stage 5 for the endpoint, auth path, multipart shape, response codes, and the correct ordering of the three uploads.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · references/skillhub-publish.md (reported line 17)May include surrounding context.

md
## Auth

Bearer token from `~/.skillhub/credentials.json` → **`user.token`**.
The file has no top-level `token` key; reading `d["token"]` yields null and produces a misleading 401/403.

## Request — multipart/form-data

Code capability 'network' detected in scripts/sync_skill_to_github.js but not covered by declared permissions.

High
Category
MCP Least Privilege
Confidence
75% confidence
Finding

The skill uses 'network' capability that is not covered by the declared 'allowed-tools' preapproval. Review the host runtime's actual enforcement policy; runtime permissions are unknown unless separately verified, and this declaration alone does not establish a permission ceiling.

Content

No source excerpt is available for this finding.

Credential Access

High
Category
Privilege Escalation
Confidence
70% confidence
Finding

Code accesses credential files (SSH keys, AWS credentials, etc.). This could indicate credential theft attempts.

Content

Scanner excerpt · scripts/sync_skill_to_github.js (reported line 11)May include surrounding context.

js
//                                [--files "SKILL.md,README.md,references/foo.md"]
//
// Environment:
//   GITHUB_TOKEN or GITHUB_PAT — a GitHub personal access token (repo scope). Required.
//   The script exits with an error if neither variable is set. Credentials are
//   consumed from the environment only — nothing credential-related is read from
//   disk, and requests go nowhere except api.github.com.

The audit criteria prescribe English SKILL.md bodies and bilingual READMEs without user opt-in or a documented justification.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

Listing these language formats under 'What It Checks' establishes a language requirement for audited skills. The README provides neither a user language choice nor a documented reason that would justify the restriction.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
50% confidence
Finding

Data is being sent to an external URL. This could be legitimate telemetry or data exfiltration. Manual review is recommended.

Content

Scanner excerpt · references/skillhub-publish.md (reported line 12)May include surrounding context.

Endpoint

text
POST https://api.skillhub.cn/api/v1/community/skills/publish

Auth

The language checklist requires an English SKILL.md body and bilingual Chinese/English README without user opt-in.

Low
Category
Output Handling
Confidence
96% confidence
Finding

The checklist prescribes specific natural languages for the skill's documentation. It neither offers a language choice nor explains a platform or locale requirement that justifies these constraints.

Content

No source excerpt is available for this finding.

The publish payload prescribes an English description without user opt-in or a documented locale requirement.

Low
Category
Output Handling
Confidence
90% confidence
Finding

The description placeholder specifies a "one-line English summary," directing the agent to generate this metadata in a particular language. The file provides neither a language choice nor a documented platform requirement that justifies this restriction.

Content

No source excerpt is available for this finding.