Back to skill

Security audit

会议纪要 · 录音成稿 · 待办提取

Security checks for vulnerabilities and agentic risk

Overview

ClawScan marks this skill suspicious because its scanners report remote service discovery and credential-handling behavior that does not clearly match its stated meeting-minutes purpose; the scan was partial, so the evidence is incomplete.

Consider not installing until the credential destination checks, network behavior, and mismatch between the stated purpose and connectivity features have been reviewed. The scan is incomplete: it covered 11 of 13 components, with partial inspection of two, so this verdict does not establish that unreported content is safe or unsafe.

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
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Findings (27)

Tainted flow: 'req' from os.environ.get (line 188, credential/environment) → urllib.request.urlopen (network output)

Critical
Category
Data Flow
Confidence
90% confidence
Finding

Credentials or environment variables flow to a network sink. This is a high-confidence indicator of credential exfiltration.

Content

Scanner excerpt · scripts/cue_connectivity_check.py (reported line 193)May include surrounding context.

python
code = None
    raw = ""
    try:
        with urllib.request.urlopen(req, timeout=args.timeout) as resp:
            code, raw = resp.status, resp.read().decode("utf-8", "replace")
    except urllib.error.HTTPError as err:
        detail = ""

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · README.en.md (reported line 88)May include surrounding context.

bash
   CATALOG="https://cuecue.cn/api/mcp-catalog"
   curl -s "$CATALOG" | python3 -c 'import json,sys
   rows=json.load(sys.stdin)["data"]["domains"]
   [print(r.get("id"), r.get("external_status"), (r.get("routing") or {}).get("url", "-")) for r in rows]'
   DOMAIN_ID="<replace with one of the ids listed by the line above>"

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · README.en.md (reported line 92)May include surrounding context.

md
rows=json.load(sys.stdin)["data"]["domains"]
   [print(r.get("id"), r.get("external_status"), (r.get("routing") or {}).get("url", "-")) for r in rows]'
   DOMAIN_ID="<replace with one of the ids listed by the line above>"
   ROUTING_URL="$(curl -s "$CATALOG" | python3 -c 'import json,sys
   rows=json.load(sys.stdin)["data"]["domains"]
   hit=[r["routing"]["url"] for r in rows if r.get("id")==sys.argv[1] and r.get("routing")]
   print(*hit,sep="\n") if hit else sys.exit("Not found: that id is not in the catalog response (or its routing has no url field) - run the line above and pick one of the ids listed")' "$DOMAIN_ID")"

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · README.md (reported line 86)May include surrounding context.

bash
   CATALOG="https://cuecue.cn/api/mcp-catalog"
   curl -s "$CATALOG" | python3 -c 'import json,sys
   rows=json.load(sys.stdin)["data"]["domains"]
   [print(r.get("id"), r.get("external_status"), (r.get("routing") or {}).get("url", "-")) for r in rows]'
   DOMAIN_ID="<把这里换成上一行列出的那个域 id>"

External Script Fetching

High
Category
Supply Chain
Confidence
90% confidence
Finding

Remote code is downloaded and executed. This bypasses code review and could introduce malicious code.

Content

Scanner excerpt · README.md (reported line 90)May include surrounding context.

md
rows=json.load(sys.stdin)["data"]["domains"]
   [print(r.get("id"), r.get("external_status"), (r.get("routing") or {}).get("url", "-")) for r in rows]'
   DOMAIN_ID="<把这里换成上一行列出的那个域 id>"
   ROUTING_URL="$(curl -s "$CATALOG" | python3 -c 'import json,sys
   rows=json.load(sys.stdin)["data"]["domains"]
   hit=[r["routing"]["url"] for r in rows if r.get("id")==sys.argv[1] and r.get("routing")]
   print(*hit,sep="\n") if hit else sys.exit("未取到:该 id 不在目录现值里(或它没有 routing 里的 url 字段)——先跑上面那行列出现场的 id 再填")' "$DOMAIN_ID")"

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The supplied code's primary function is channel discovery and authenticated connectivity testing, rather than recording transcription or source-anchored meeting-minute generation. It neither processes audio nor produces any of the declared five sections. Although connectivity checks could support a larger workflow, this chunk exposes a general-purpose external-service diagnostic without any demonstrated connection to meeting processing. Its default network path performs an anonymous catalog GET even without --post; the authenticated request is optional and initially restricted to HTTPS destinations matching authorized host suffixes. These safeguards limit credential use but do not resolve the undeclared capability and resource-access mismatch.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
94% confidence
Finding

The supplied chunk performs service discovery and remote tool enumeration, with no demonstrated recording processing, transcription, timestamp anchoring, or meeting-minutes generation. Such discovery could support a larger implementation, but this chunk does not establish any connection between the selected service and the declared meeting-minutes workflow. It also sends an API credential to the catalog-provided routing URL; this is observable behavior, not proof of malicious credential theft.

Content

No source excerpt is available for this finding.

Description-Behavior Mismatch

High
Category
MCP Tool Poisoning
Confidence
93% confidence
Finding

该代码块实际执行的是通用远程服务发现与工具枚举,没有读取录音、转写、建立时间锚或生成会议纪要。工具发现可能是更大工作流的辅助步骤,但本片段没有显示所选服务与会议处理存在关联,且会向动态选定的地址发送认证密钥。因此,按所提供代码判断,其可见行为与声明用途存在实质差异;这并不单独证明存在恶意行为。

Content

No source excerpt is available for this finding.

Automatic redirects can bypass the promised credential destination checks.

High
Category
Excessive Agency
Confidence
98% confidence
Finding

The destination guard checks only the initial URL, while urllib's default redirect handler can follow a POST's 301, 302, or 303 response and copy its Authorization header onto the redirected GET. An allowed endpoint that is compromised or exposes an attacker-controlled redirect could therefore send the API key to an unauthorized host or an HTTP destination without another guard check.

Content

No source excerpt is available for this finding.

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

Medium
Category
MCP Rug Pull
Confidence
86% confidence
Finding

The documented installation command invokes an unpinned npm executable through npx skills and installs skill content without an immutable source reference. Although this is an installation example rather than an MCP server declaration, following it can execute changed or compromised upstream code with the user's privileges; its presence in historical documentation limits immediate exposure but does not eliminate the supply-chain risk.

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 44)May include surrounding context.

bash
# 1. Put it in the skills directory
mkdir -p ~/.workbuddy/skills
cp -r meeting-minutes-cn ~/.workbuddy/skills/

# 2. Copy the card template and fill it following section 1

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 42)May include surrounding context.

bash
# 1. 放到技能目录
mkdir -p ~/.workbuddy/skills
cp -r meeting-minutes-cn ~/.workbuddy/skills/

# 2. 复制卡模板,按 §一 三步填

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The command executes an unpinned npm CLI and installs skills from a mutable repository reference. Although this is not itself an MCP server declaration, compromised upstream updates could execute code or introduce malicious agent instructions during installation.

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 one skill limits installation scope but does not pin either the executing npm CLI or the downloaded skill revision. An upstream compromise could therefore introduce code or agent instructions with access to sensitive meeting workflows.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
98% confidence
Finding

The shell example attaches the API key to a URL taken directly from catalog data without validating its scheme or host. A compromised catalog could direct the bearer token to an attacker-controlled or plaintext endpoint; the separately described script's safeguards do not protect this shell command.

Content

Scanner excerpt · README.en.md (reported line 97)May include surrounding context.

md
hit=[r["routing"]["url"] for r in rows if r.get("id")==sys.argv[1] and r.get("routing")]
   print(*hit,sep="\n") if hit else sys.exit("Not found: that id is not in the catalog response (or its routing has no url field) - run the line above and pick one of the ids listed")' "$DOMAIN_ID")"
   [ -n "$ROUTING_URL" ] || { echo "Fetch failed: ROUTING_URL is empty - run the line above, then set DOMAIN_ID to one of the ids listed" >&2; exit 1; }
   curl -s -X POST "$ROUTING_URL" \
     -H "Authorization: Bearer $CUE_API_KEY" \
     -H "Accept: application/json, text/event-stream" \
     -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' \

The parsing workflow lacks an explicit warning that meeting recordings are sent to a third-party service.

Medium
Category
Output Handling
Confidence
89% confidence
Finding

The workflow requests consent for parsing expenditure but does not clearly disclose that recording content will be sent to an external processing service. References to local files and a local Bridge component may leave users unaware of the confidentiality implications for corporate recordings.

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 package's own installation command leaves both the npm installer and repository content unpinned. Installation can consequently acquire different code or agent instructions without a corresponding change to this documentation.

Content

No source excerpt is available for this finding.

The recording-parsing workflow lacks an explicit warning that meeting content is sent to a third-party service.

Medium
Category
Output Handling
Confidence
91% confidence
Finding

The instructions discuss a local Bridge, service credentials, and spending approval without explicitly warning that meeting content is transferred to an external processor. Users may authorize parsing costs without understanding that confidential recordings leave their local environment.

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 command executes an unpinned npm CLI and retrieves skill content from a mutable repository reference. Although the match is an installer rather than an MCP server declaration, upstream compromise could introduce executable code or malicious 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

Installing only cue-omni-reader reduces scope but leaves the executing CLI and downloaded skill revision unpinned. A compromised upstream update could affect an agent handling sensitive recordings.

Content

No source excerpt is available for this finding.

External Transmission

Medium
Category
Data Exfiltration
Confidence
98% confidence
Finding

The shell command sends the bearer credential to a catalog-supplied URL after checking only that the value is nonempty. Malicious catalog data could redirect the key to an unauthorized host or HTTP endpoint, and the separate helper's documented protections do not apply to this shell path.

Content

Scanner excerpt · README.md (reported line 95)May include surrounding context.

md
hit=[r["routing"]["url"] for r in rows if r.get("id")==sys.argv[1] and r.get("routing")]
   print(*hit,sep="\n") if hit else sys.exit("未取到:该 id 不在目录现值里(或它没有 routing 里的 url 字段)——先跑上面那行列出现场的 id 再填")' "$DOMAIN_ID")"
   [ -n "$ROUTING_URL" ] || { echo "取值失败:ROUTING_URL 为空——先跑上面那行列出现场 id,再把 DOMAIN_ID 换成其中之一" >&2; exit 1; }
   curl -s -X POST "$ROUTING_URL" \
     -H "Authorization: Bearer $CUE_API_KEY" \
     -H "Accept: application/json, text/event-stream" \
     -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' \

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

Medium
Category
MCP Rug Pull
Confidence
94% confidence
Finding

The self-installation command does not pin the npm installer or the repository revision containing the skill. This permits installation behavior and agent instructions to change without any change to the documented command.

Content

No source excerpt is available for this finding.

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

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 audio-processing workflow lacks a warning about third-party handling of meeting content.

Medium
Category
Output Handling
Confidence
86% confidence
Finding

The workflow directs recordings into a processing channel without clearly stating whether meeting content leaves the local environment or requiring approval before remote submission. Enterprise recordings can contain confidential business information and personal data, so this missing disclosure and consent checkpoint can enable unintended third-party exposure; the supplied text does not establish that an actual disclosure has occurred.

Content

No source excerpt is available for this finding.

A document-wide price ban rejects legitimate meeting content despite the skill's faithful-transcription purpose.

Medium
Category
Excessive Agency
Confidence
95% confidence
Finding

The price prohibition scans the entire document, including verbatim meeting content, rather than distinguishing service-pricing claims from recorded financial discussions. In a faithful-transcription workflow, this can block legitimate minutes or pressure the agent to omit material financial details to obtain a passing result, undermining record integrity.

Content

No source excerpt is available for this finding.