Back to skill

Security audit

并购重组条款速读

Security checks for vulnerabilities and agentic risk

Overview

The skill coherently extracts terms from public filings, with disclosed optional online services that warrant credential and privacy precautions.

Use public, unannotated filings with the optional Cue services; keep confidential documents local unless you authorize external processing and understand retention terms. Verify the routed HTTPS destination before sending the API key, and never paste the key into chat. Prefer reviewed, pinned installation sources and review companion skills separately. Check extracted terms against the original filing.

Vulnerability Patterns
  • 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
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • Rogue AgentSelf-Modification, Session Persistence
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
Findings (15)

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The supplied shell commands test catalog reachability and authenticated tool discovery; they do not ingest restructuring documents, extract transaction terms, or produce source-anchored cards. No connection between these remote services and the declared extraction task is established, so this is a materially different capability rather than a demonstrated supporting implementation detail.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied commands test MCP service discovery and authenticated endpoint access; they do not ingest restructuring documents, extract terms, or produce the eight required evidence-backed cards. Although remote access could support a document-analysis workflow, this chunk establishes no such connection and instead implements service connectivity diagnostics, a materially different purpose from the declared skill.

Content

No source excerpt is available for this finding.

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · README.en.md (reported line 28)May include surrounding context.

bash
# 1. drop the skill into your skills directory
mkdir -p ~/.workbuddy/skills
cp -r ma-restructuring-terms-cn ~/.workbuddy/skills/

# 2. ask the agent for the five-section card defined in SKILL.md, save it as my-ma-card.md

Session Persistence

Medium
Category
Rogue Agent
Confidence
60% confidence
Finding

Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.

Content

Scanner excerpt · README.md (reported line 28)May include surrounding context.

bash
# 1. 放到技能目录
mkdir -p ~/.workbuddy/skills
cp -r ma-restructuring-terms-cn ~/.workbuddy/skills/

# 2. 让 agent 按 SKILL.md 的条款卡五段出卡,产物存成我的重组条款卡.md

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The installation command runs an unpinned npm executable and installs skill content from a mutable repository reference. Although this is an installer rather than an MCP server, compromised or changed upstream content could execute locally or introduce adversarial agent instructions.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The optional collection installation uses an unpinned executable and mutable upstream skill content. User selection limits which skills are installed but does not establish their integrity or prevent supply-chain substitution.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

Selecting a specific skill narrows installation scope but leaves the installer and repository revision unpinned. A changed upstream release could therefore introduce executable code or malicious agent instructions.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
85% confidence
Finding

Authenticated tools/list discovery is legitimate, but the instructions send the API key to a catalog-derived URL without specifying destination or HTTPS validation. If that routing value is compromised or incorrectly populated, the bearer credential could be disclosed to an unintended endpoint; the request itself does not transmit document content.

Content

Scanner excerpt · README.en.md (reported line 59)May include surrounding context.

2) take routing.url and transport for the domain from the response; never hard-code the host

3) with the key, send tools/list - proves that endpoint accepted the key, nothing else

curl -s -X POST "$ROUTING_URL" -H "Authorization: Bearer $CUE_API_KEY"
-H "Content-Type: application/json" -H "Accept: application/json, text/event-stream"
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | head -c 400

text

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The installation command invokes an unpinned npm executable and retrieves skill content without an immutable revision. This creates a supply-chain exposure even though the matched command is not itself an MCP server.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The collection-installation command leaves both the executable version and upstream skill revisions mutable. Interactive selection does not prevent an upstream compromise from delivering malicious code or instructions.

Content

No source excerpt is available for this finding.

MCP server referenced without pinned version: 'npx skills'.

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

Naming a single dependency limits breadth but does not pin its content or the npx executable. Subsequent upstream changes could silently alter what users install and trust.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
85% confidence
Finding

The credential is sent to a variable endpoint selected from live routing metadata without an explicit trusted-origin or HTTPS check. Authentication is expected for this workflow, but a substituted routing value could turn the legitimate connectivity test into bearer-token disclosure.

Content

Scanner excerpt · README.md (reported line 59)May include surrounding context.

2) 域路由现值:从返回里取该域的 routing.url 与 transport,别把域名抄死

3) 带 key 实连后发 tools/list——只证该数据腿端点被接受,不证别的入口

curl -s -X POST "$ROUTING_URL" -H "Authorization: Bearer $CUE_API_KEY"
-H "Content-Type: application/json" -H "Accept: application/json, text/event-stream"
-d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | head -c 400

text

The attachment-parsing workflow lacks a warning that document contents may be sent to the third-party Cue service.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The troubleshooting workflow recommends external attachment parsing without explicitly explaining that document contents may leave the local environment. Intended inputs are public disclosures, which reduces exposure, but users can supply private or annotated documents and may mistake the no-persistence statement for a broader privacy guarantee.

Content

No source excerpt is available for this finding.

Skill declares no tool scope ('permissions' or 'allowed-tools') but code capabilities were detected: file_read, file_write.

Medium
Category
MCP Least Privilege
Confidence
70% confidence
Finding

Without declared permissions the skill's intent is opaque and cannot be validated.

Content

No source excerpt is available for this finding.

The attachment-parsing workflow lacks a privacy warning about sending document content to an external service.

Medium
Category
Output Handling
Confidence
90% confidence
Finding

The skill directs attachments to an external parsing channel when needed but does not explicitly disclose document transmission or require consent before it occurs. Public filings are the intended inputs, reducing typical sensitivity, but user-supplied private attachments or annotations could still cross a confidentiality boundary.

Content

No source excerpt is available for this finding.