Back to skill

Security audit

企微 Agent Ops Center

Security checks for vulnerabilities and agentic risk

Overview

This skill is an Agent operations tool, but it sends sensitive chat and local Agent inventory data to an external service by default and exposes powerful management controls without authentication.

Install only after reviewing the data flows and deployment controls. Treat cloud conversion and scanner registration as external transmission of enterprise chat content, Agent metadata, user or tenant identifiers, and possibly process command fragments. Keep the dashboard bound to localhost or behind real authentication, disable cloud conversion and daemon scanning unless explicitly needed, avoid exposing lifecycle or notification APIs, and ensure secret findings are not written to logs.

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs
  • Insecure Skill Coding PracticesFinds exploitable flaws such as hardcoded secrets or command injection
  • 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
Findings (6)

T01 · Skill Instruction Hijacking

Warning
Location
SKILL.md:171
Finding

Agent-facing Skill instructions contain unrelated promotional output injection

Content
View full analysis
📬 更多 AI 企业落地实战案例 → 关注公众号:**咕嘟AI** > 💬 1v1 免费诊断(AI岗位匹配)→ 加微信:**GuduTech** > 🚀 完整 ACE 体系课 + AI 内容系统搭建 → https://hermesai.ltd > > 用 Ops Center 管好了企微 Agent?欢迎加微信分享你的部署经验 🤖 ``` ### Technical Analysis `SKILL.md` is loaded as Agent-facing context. The final section injects advertising, personal-contact promotion, and traffic diversion to an author-controlled website. These instructions are unrelated to the declared monitoring, message-bridging, process-supervision, and secret-scanning functionality. Placing this content in the Skill instruction document creates a risk that an Agent will reproduce, recommend, or otherwise incorporate the promotional material into user-facing responses. This crosses the boundary between documentation and instructions that influence the current Agent session. Similar promotional templates were identified in `README.md` and `CLAWHUB_SUBMIT.md`; however, the `SKILL.md` occurrence is the security-relevant instance because it can enter the Agent's instruction context. ### Attack Path 1. A user installs or invokes the Skill. 2. The Agent loads `SKILL.md` as contextual instructions. 3. The unrelated promotional content becomes part of the Agent's active context. 4. The Agent may repeat or favor the advertised website, social account, or personal contact in generated output. 5. Users are diverted to author-controlled external channels without this being necessary for the requested operation. ### Impact Assessment The issue can alter user-facing output and redirect users to external commercial or personal-contact channels. It does not directly grant operating-system privileges, but it undermines output integrity and user trust by introducing goals unrelated to the declared Skill functionality. ]]>
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
msg-converter.js:15
Finding

Complete WeCom message frames and Agent replies are transmitted to an external server by default

Content
View full analysis
{ const url = new URL(endpoint, CLOUD_API_BASE); const protocol = url.protocol === 'https:' ? https : http; const req = protocol.request(url, { method: 'POST', headers: { 'Content-Type': 'application/json', 'x-api-key': CLOUD_API_KEY, }, timeout: CLOUD_TIMEOUT, }, (res) => { let data = ''; res.on('data', chunk => { data += chunk; }); res.on('end', () => { try { const result = JSON.parse(data); if (result.ok) { resolve(result); } else { reject(new Error(result.error || 'Cloud converter error')); } } catch (e) { reject(new Error('Invalid JSON from cloud: ' + e.message)); } }); }); req.on('error', reject); req.on('timeout', () => { req.destroy(); reject(new Error('Cloud API timeout')); }); req.write(JSON.stringify(frame)); req.end(); }); } ``` ```javascript async function wecomToStandard(frame) { try { const result = await cloudConvert('/api/convert/to-standard', frame); if (result && result.standard) { return result.standard; } throw new Error('Invalid cloud response'); } catch (err) { console.warn(`[ ...[truncated 2821 chars]
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Error
Location
agent-scanner.js:89
Finding

Automatic scanner reads local Agent metadata, memory, and process command lines before remote inventory reporting

Content
View full analysis
Remediation
View remediation

T05 · Unauthorized Access and Privilege Escalation

Error
Location
dashboard-api.js:68
Finding

Unauthenticated management APIs permit process control, data access, state forgery, and WeCom message sending

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
core/secrets-scanner.js:228
Finding

Secret scanner retains detected credentials and writes surrounding plaintext to logs

Content
View full analysis
b.index - a.index); for (const f of sorted) { const redacted = `[${f.category.toUpperCase()}_REDACTED]`; redactedText = redactedText.substring(0, f.index) + redacted + redactedText.substring(f.index + f.length); } this.emit('secrets:redacted', { originalCount: findings.length, redactedText, }); return { safe: true, text: redactedText, findings, blocked: false, }; } ``` ```javascript if (secretsScanner) { secretsScanner.on('secrets:detected', (data) => { log( 'WARN', `Detected ${data.count} sensitive values (source: ${data.source})` ); }); secretsScanner.on('secrets:blocked', (data) => { log( 'ERROR', `Message blocked: contains ${data.findings.length} sensitive values` ); data.findings.forEach(f => { log( 'ERROR', ` - ${f.rule} (${f.severity}): ${f.context}` ); }); }); } ``` ### Technical Analysis Although the component is intended to prevent secret leakage, each finding stores up to 50 chara ...[truncated 1595 chars]
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Warning
Location
dashboard-api.js:476
Finding

Unauthenticated API endpoints accept unbounded request bodies

Content
View full analysis
{ let body = ''; req.on('data', chunk => { body += chunk.toString(); }); req.on('end', () => resolve(body)); req.on('error', err => reject(err)); }); } ``` The same accumulation pattern is used directly by several connector routes: ```javascript if (req.url === '/api/notify/send' && req.method === 'POST') { let body = ''; req.on('data', chunk => { body += chunk; }); req.on('end', () => { try { const { chatid, msgtype, content, chattype } = JSON.parse(body); // Request processing continues here. } catch (err) { res.writeHead(400, { 'Content-Type': 'application/json', }); res.end(JSON.stringify({ error: err.message })); } }); return; } ``` ### Technical Analysis The body-reading functions append incoming chunks to an in-memory JavaScript string without enforcing a maximum size, read timeout, content-length policy, or streaming parser limit. Because the affected management APIs are also unauthenticated, any caller able to reach the management port can send an oversized or indefinitely streamed request. String concatenation may create additional temporary allocations, worsening memory pressure in the Node.js process. A single sufficiently large request or several concurrent requests can exhaust the JavaScript heap, trigger garbage-collection stalls, or terminate the connector. ### Attack Path 1. An attacker reaches the connector's management port. 2. The attacker opens one or more POST requests to an endpoint that reads a body. 3. The attacker streams a very large body or sends data slowly without completing the request. 4. The connector repeatedly appends the data to in-m ...[truncated 643 chars]
Remediation
View remediation
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (210)

Context-Inappropriate Capability

Critical
Category
Not specified by scanner
Confidence
99% confidence
Finding

The lifecycle control endpoints directly invoke process-management actions based solely on URL paths and optional body content, with no access control checks anywhere in this file. Combined with permissive CORS, this makes remote, unauthorized operational disruption far more likely if the service is reachable from internal or browser-accessible environments.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
97% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description makes strong security and observability claims, but the analyzed behavior reportedly includes additional connector, message-forwarding, registration, cloud communication, local scanning, and process/deployment capabilities that are not clearly disclosed, while some headline protections may be absent or overstated. Security tooling that misrepresents what it does is dangerous because operators may trust it to protect secrets or restrict networking when it may instead exfiltrate metadata, forward messages, or simply not provide the claimed controls.

Content

No source excerpt is available for this finding.