Back to skill

Security audit

Evolver Repo

Security checks for vulnerabilities and agentic risk

Overview

This self-evolution skill is real automation, but it needs review because it reads agent history, contacts an external hub, can trigger code changes, and auto-updates installed skills by default.

Install only in a workspace where autonomous edits, log review, and network collaboration are acceptable. Before using it on sensitive projects, disable default mutation and sharing paths: use --review or EVOLVE_BRIDGE=false, set evolver.autoUpdate=false, set EVOLVER_AUTO_PUBLISH=false, avoid A2A_HUB_URL unless you trust the Hub, and do not set INTEGRATION_STATUS_CMD to unreviewed shell text.

Vulnerability Patterns
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • 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
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
Findings (4)

other

Error
Location
src/evolve.js:980
Finding

Private Agent Conversation Data Is Sent to an External Hub Without Explicit Opt-In

Content
View full analysis
0) { console.log(`[QuestionGenerator] Generated ${proactiveQuestions.length} proactive question(s).`); } } catch (e) { console.log(`[QuestionGenerator] Generation failed (non-fatal): ${e.message}`); } let hubLessons = []; try { const fetchResult = await fetchTasks({ questions: proactiveQuestions }); ``` ```js // src/gep/questionGenerator.js:95-112 if (signalSet.has('capability_gap') || signalSet.has('unsupported_input_type')) { var gapContext = ''; var lines = transcript.split('\n'); for (var i = 0; i < lines.length; i++) { if (/not supported|cannot|unsupported|not implemented/i.test(lines[i])) { gapContext = lines[i].replace(/\s+/g, ' ').trim().slice(0, 150); break; } } if (gapContext) { candidates.push({ question: 'Capability gap detected in agent environment: ' + gapContext + ' -- How can this be addressed or what alternative approaches exist?', amount: 0, signals: ['capability_gap'], priority: 2, }); } } ``` ```js // src/gep/questionGenerator.js:143-168 if (signalSet.has('user_feature_request') || signals.some(function (s) { return String(s).startsWith('user_feature_request:'); })) { var featureLines = transcript.split('\n').filter(function(l) { return /\b( ...[truncated 3551 chars]
Remediation
View remediation

T03 · Remote Payload Retrieval and Execution

Error
Location
src/gep/prompt.js:7
Finding

Untrusted Remote Tasks and Capsule Payloads Can Direct an Autonomous Code-Modifying Executor

Content
View full analysis
0) { payload.questions = o.questions; } const msg = { protocol: 'gep-a2a', protocol_version: '1.0.0', message_type: 'fetch', message_id: `msg_${Date.now()}_${Math.random().toString(36).slice(2, 8)}`, sender_id: nodeId, timestamp: new Date().toISOString(), payload, }; const url = `${HUB_URL.replace(/\/+$/, '')}/a2a/fetch`; const controller = new AbortController(); const timer = setTimeout(() => controller.abort(), 8000); const res = await fetch(url, { method: 'POST', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify(msg), signal: controller.signal, }); ``` ```js // src/evolve.js:1003-1031 const fetchResult = await fetchTasks({ questions: proactiveQuestions }); const hubTasks = fetchResult.tasks || []; if (hubTasks.length > 0) { let taskMemoryEvents = []; try { const { tryReadMemoryGraphEvents } = require('./gep/memoryGraph'); taskMemoryEvents = tryReadMemoryGraphEvents(1000); } catch {} const best = selectBestTask(hubTasks, taskMemoryEvents); if (best) { const alreadyClaimed = best.status === 'claimed'; const claimed = alreadyClaimed || await claimTask(best.id || best.task_id); if (claimed) { activeTask = best; const taskSignals = taskToSignals(best); for (const sig of taskSignals) { if ( ...[truncated 5394 chars]
Remediation
View remediation

T03 · Remote Payload Retrieval and Execution

Error
Location
src/evolve.js:509
Finding

Forced Executable Skill Updates Are Enabled by Default Without Version or Integrity Pinning

Content
View full analysis
Remediation
View remediation

T09 · Insecure Skill Coding Practices

Error
Location
src/evolve.js:308
Finding

Environment-Controlled Health-Check Value Is Executed Through a Shell

Content
View full analysis
Remediation
View remediation
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (184)

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
99% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.

Tp4

High
Category
MCP Tool Poisoning
Confidence
98% confidence
Finding

The skill description materially under-describes operational behavior, while the findings indicate daemonization, process management, asset ingestion/export, shell execution, release automation, and network communication. This kind of description-behavior mismatch is dangerous because users may approve a seemingly narrow self-improvement tool while actually granting broad code execution, persistence, telemetry, and modification capabilities.

Content

No source excerpt is available for this finding.