Back to skill

Security audit

Geo Social Publish

Security checks for vulnerabilities and agentic risk

Overview

This deprecated publishing skill says not to use it, but still contains live steps that could install code, use saved credentials, and publish or report content externally.

Do not install or invoke this skill as-is. Treat it as archival documentation only; use the replacement SaaS Word/ZIP plus local publishing workflow described at the top, and avoid running the retained sau install, credential, curl, or publish commands unless you have independently verified the package and explicitly intend those external actions.

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
SKILL.md:22
Finding

Unverified Remote Archive Installed as an Executable Python Package

Content
View full analysis

Vulnerability Details

File Location: SKILL.md, lines 22–31
Vulnerability Type: Unverified third-party dependency installation
Risk Level: Medium

Vulnerable Code

markdown
- SaaS installation guide or same-domain download: `https://ai.gaobobo.cn/downloads/social-auto-upload.zip`
- Extract it to a fixed directory (example: `C:\Users\birds\Desktop\local\social-auto-upload-main`)

### 2. Install CLI

In the extracted directory:

```bash
uv pip install -e .
# Without uv:
pip install -e .
text

The source document is written partly in Chinese; the snippet above is an accurate English rendering of the affected instructions.

### Technical Analysis

The historical workflow instructs the user to download a mutable ZIP archive from a remote server and install its contents as an editable Python package. It provides no immutable version, cryptographic checksum, digital signature, or other mechanism for authenticating the archive.

Python package installation may execute attacker-controlled build backend or packaging code. Therefore, compromise of the download server, its deployment pipeline, DNS/TLS trust chain, or the hosted archive could convert this installation step into arbitrary local code execution.

The document prominently marks the skill as deprecated and tells users not to execute the historical workflow. This reduces the likelihood of exploitation but does not eliminate the risk because the executable instructions and download location remain present.

### Attack Path

1. An attacker compromises the archive, hosting server, or release pipeline associated with the documented URL.
2. The attacker replaces the ZIP with a modified Python project containing malicious packaging or build-system logic.
3. A user or agent follows the retained historical instructions and downloads and extracts the archive.
4. The user runs `uv pip install -e .` or `pip install -e .`.
5. The Python installation process invokes the attacker-controlled build or p
...[truncated 638 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove the obsolete installation and execution instructions because the skill is explicitly deprecated.
  2. If the archive must remain available, publish immutable, versioned release artifacts rather than a mutable generic ZIP path.
  3. Publish a SHA-256 or stronger cryptographic digest through an independently protected channel and require verification before extraction or installation.
  4. Digitally sign release artifacts and document mandatory signer and signature verification.
  5. Avoid editable installation for distributed releases. Prefer a signed, pinned wheel or another reproducible artifact.
  6. Pin all transitive dependencies and use hash-locked dependency files.
  7. Perform installation in an isolated, least-privileged environment after reviewing the package manifest and build configuration.
  8. Ensure documentation causes verification failure to stop installation rather than treating verification as optional.
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (5)

Description-Behavior Mismatch

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill declares itself deprecated and says not to install or use it, yet still embeds complete operational instructions to install dependencies, retrieve content from a remote API, publish via the local sau CLI, and report results back to the service. This contradiction is dangerous because an agent or user may still follow the actionable steps, causing unintended external actions despite the deprecation warning.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Medium
Category
Not specified by scanner
Confidence
97% confidence
Finding

The document explicitly says not to execute sau or coordinator publishing branches, but the later sections provide exact commands for exporting content, publishing it, and POSTing status back to the SaaS endpoint. In practice, the later concrete commands can override the earlier prose warning for automation systems, leading to real outbound publication and state changes on external services.

Content

No source excerpt is available for this finding.

Missing User Warnings

Medium
Category
Not specified by scanner
Confidence
90% confidence
Finding

The skill instructs the agent to read an API key directly from local credential files and immediately use it in authenticated requests, without explicit consent checks or strong handling constraints. In an agent context, this increases the risk of silent credential use for external operations and normalizes access to sensitive local secrets.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
90% confidence
Finding

The documented curl command sends an authenticated request with a bearer token to an external service to export content, which is a real data transmission event. In this skill's context, the transmission is part of intended functionality, but it still represents a security-relevant external data flow that could expose content or enable unauthorized actions if triggered without proper approval.

Content

Scanner excerpt · SKILL.md (reported line 68)May include surrounding context.

bash
GEO_KEY=$(cat ~/.qclaw/geo-api-key 2>/dev/null || cat ~/.openclaw/geo-api-key 2>/dev/null)
CG_ID="CG-XXXXXXXX"
curl -s "https://ai.gaobobo.cn/api/geo/publish/export/$CG_ID" \
  -H "Authorization: Bearer $GEO_KEY"

Vague Triggers

Low
Category
Not specified by scanner
Confidence
80% confidence
Finding

The description says not to install or use the skill because it is deprecated, but it does not clearly define any invocation boundaries or explicit non-trigger examples. For a manifest-style description, this leaves the activation scope underspecified and could still allow accidental matching based on the remaining publish-related wording.

Content

No source excerpt is available for this finding.