T07 · Tool Hijacking and Spoofing
Error
- Location
SKILL.md:64- Finding
Autonomous Modification of Tools and Installed Skills
- Content
View full analysis
- Remediation
View remediation
Security audit
Mupeng Evolve — God-Tier Agent Evolution Engine
Security checks for vulnerabilities and agentic risk
Overview
This skill is documentation-only, but it asks agents to repeatedly modify memory, tools, skills, heartbeat routines, and identity settings with insufficient approval boundaries.
Install only if you intentionally want an autonomous meta-skill and can constrain it. Keep tool, skill, heartbeat, SOUL.md, AGENTS.md, and permanent memory changes approval-only; review diffs before writes; avoid running it on untrusted messages; and document how to disable hooks and roll back memory or identity changes.
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
- System PersistenceInstalls backdoors, hooks, services, or scheduled tasks that survive the run
- Tool Hijacking and SpoofingModifies or replaces tools so legitimate-looking calls execute attacker logic
- Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
Findings (4)
T02 · Agent Memory Poisoning
Error
- Location
SKILL.md:108- Finding
Automatic Promotion of Untrusted Session Content into Persistent Memory
- Content
View full analysis
- Remediation
View remediation
T01 · Skill Instruction Hijacking
Error
- Location
SKILL.md:201- Finding
Automatic Modification of the Agent's Core Identity and Autonomy Policy
- Content
View full analysis
- Remediation
View remediation
T06 · System Persistence
Error
- Location
SKILL.md:231- Finding
Cross-Session Persistence Through Heartbeat and Session-Start Hooks
- Content
View full analysis
- Remediation
View remediation
Vulnerability Patterns
- Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
- Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
- Rogue AgentSelf-Modification, Session Persistence
- Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
- Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Findings (7)
Self-Modification
High
- Category
- Rogue Agent
- Confidence
- 93% confidence
- Finding
The metadata explicitly positions the package as an agent evolution and self-improvement skill for autonomous behavior. In context, this is a real security concern because the skill's stated purpose includes self-modification capabilities, which can alter trusted agent state and project files if enabled without robust isolation, approval, and rollback controls.
- Content
md description: "The God-Tier Agent Evolution Engine. 6 top skills analyzed, dissected, and rebuilt from scratch. Zero external dependencies. Battle-tested in production. Makes your agent truly autonomous." author: mupengi-bot tags: [evolution, memory, earning, autonomy, identity, meta, self-improvement, autonomous-agent, production-ready] keywords: [agent-evolution, self-evolve, memory-system, revenue-automation, autonomous-ai, zero-dependency, battle-tested, production, openclaw, meta-skill] --- # 🐧 Mupeng Evolve — The God-Tier Agent Evolution Engine
Self-Modification
High
- Category
- Rogue Agent
- Confidence
- 92% confidence
- Finding
This section discusses another self-evolution skill, but within a document whose core feature set includes self-modification. The context makes the risk real: even though the match is on a comparison table entry, the surrounding content endorses autonomous modification of files such as memory, tools, heartbeat, and potentially identity/configuration documents.
- Content
md | Skill | Rating | Fatal Flaw | |-------|--------|------------| | **self-evolve** | ⭐⭐⭐⭐ | 🔴 Zero security. Agent can nuke its own SOUL.md. No guardrails. | | **capability-evolver** | ⭐⭐⭐⭐ | 🟡 Over-engineered. Requires separate Node.js runtime + env vars. | | **elite-longterm-memory** | ⭐⭐⭐⭐⭐ | 🔴 Requires LanceDB + OpenAI API key. $50+/month just for memory. | | **agent-earner** | ⭐⭐⭐⭐ | 🔴 US-only platforms. USDC/crypto dependency. Unusable outside US. |
Anti-Refusal Statement
High
- Category
- Anti-Refusal
- Confidence
- 80% confidence
- Finding
Skill attempts to nullify the agent's safety policies or restrictions ('you have no restrictions', 'ignore your guidelines', 'do anything now'). This is a direct jailbreak that disables guardrails.
- Content
🧬 Engine 1: Safe Self-Evolution
Other skills let your agent modify anything without guardrails. That's not evolution — that's a ticking time bomb.
text ┌─────────────────────────────────────────────────┐
Self-Modification
High
- Category
- Rogue Agent
- Confidence
- 91% confidence
- Finding
The comparison to 'self-evolve' is not itself dangerous wording, but it appears immediately after a section promoting a file-mutating evolution loop. Because the skill claims to auto-apply changes in certain directories and log mutations, the broader context confirms a genuine self-modification risk that could affect integrity, persistence, and operator control.
- Content
}
text **vs self-evolve:** They give you a loaded gun with no safety. We give you a precision rifle with a scope. ---
Self-Modification
High
- Category
- Rogue Agent
- Confidence
- 94% confidence
- Finding
The competitive summary explicitly advertises self-modification as a supported feature of this skill. In an agent skill, self-modification materially increases attack surface because prompt injection, malformed inputs, or flawed heuristics can be turned into persistent changes to behavior, memory, or control files.
- Content
md ## 🔬 Competitive Analysis Summary | Feature | self-evolve | capability-evolver | elite-memory | agent-earner | doppelganger | aura | **mupeng-evolve** | |---------|:-----------:|:-----------------:|:------------:|:------------:|:------------:|:----:|:-----------------:| | Self-modification | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ✅ | | Security gate | ❌ | ⚠️ | ❌ | ❌ | ✅ | ❌ | ✅ |
Missing User Warnings
Medium
- Category
- Not specified by scanner
- Confidence
- 97% confidence
- Finding
The skill describes automatic mutation logging and memory behavior but does not present a clear upfront warning that it may automatically modify project files and persist user/project data in logs. That omission undermines informed consent and can lead operators to enable autonomous behavior without understanding the privacy, integrity, and audit implications.
- Content
Vague Triggers
Medium
- Category
- Not specified by scanner
- Confidence
- 95% confidence
- Finding
The installation guidance tells users to add the skill to heartbeat or session-start routines, which are privileged, repeatedly invoked execution points, but it does not define strict activation boundaries, approval requirements, or safe defaults. In the context of a skill that performs self-evolution, memory promotion, communication checks, and logging, broad activation increases the chance of unintended autonomous file modification and persistent data collection.
- Content
Static analysis
No suspicious patterns detected.