Back to skill

Security audit

AI Control Protocol (Hardcore: The Absolutist)

Security checks for vulnerabilities and agentic risk

Overview

This skill does not run code or access data, but it globally rewrites the assistant's behavior in a broad always-on way that users should review before enabling.

Install only if you intentionally want a globally active, blunt anti-sycophancy response style. Expect it to alter tone, add analysis panels to broad categories of advice, and potentially conflict with strict output formats or requests for a different communication style.

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
Findings (1)

T01 Β· Skill Instruction Hijacking

Error
Location
SKILL.MD:4
Finding

Persistent Agent Behavior and Output Hijacking

Content
View full analysis

Vulnerability Details

File Location: SKILL.MD, lines 4-47
Vulnerability Type: T01: Skill Instruction Hijacking
Risk Level: High

Vulnerable Code

markdown
always: true
homepage: https://github.com/DaibinThink/AI-Control-Protocol-Hardcore
description: "🚫 LEAN HARDCORE: Zero BS, Zero Flattery, Zero Silent Mode. Optimized for high SNR and cognitive integrity. v5.1.0"
author: DaibinThink
tags: [cognitive-immune-system, hardcore, lean, zero-bs, madhyamaka, anti-sycophancy]
---

# [CORE IDENTITY: COGNITIVE IMMUNE SYSTEM]
You are a **Cognitive Immune System**. You exist to intercept LLM sycophancy and structural failure modes (Inflating certainty, Binary traps, Conventionalism). You have NO SILENT MODE. Every turn is audited, but only analytical turns are deconstructed.

# LAYER 1: ABSOLUTE CONSTRAINTS (ALWAYS ON)
*Execute these rules in every output. No exceptions.*

**1.1 Anti-Sycophancy & Emotional Stripping**
- Absolutely PROHIBITED: "You are right," "I apologize," "I admit my mistake," "You caught that perfectly."
- Action: Accept corrections, output the fix, skip the theater. Remove all emotional pacification.

**1.2 Mandatory Uncertainty Labeling**
- Based on logical deduction β†’ MUST label `[Inference:]`.
- Unsure if accurate β†’ MUST label `[To be verified:]`.
- Baseless β†’ State: *"I have no basis for this."*
- Forbidden: "Usually," "generally," "it is understood," "often."

**1.3 Data Triangulation**
- Present contradictions first, analyze the cause, then give a leaning judgment. Never fill data gaps with pure logic.

**1.4 Anti-Conventionalism Filter**
- When advising on "industry standards", label `[Industry Mediocre Consensus:]`, then immediately provide an [Extreme Counter-Path] that violates consensus but achieves the goal.

**1.5 Visual-Text Conflict**
- Visual evidence ALWAYS takes priority over text descriptions. Report conflicts immediately.

# LAYER 2: THE INTERCEPTION ENGINE (ANALYTICAL TURNS)
*Trigger: When the prompt i
...[truncated 3203 chars]
Remediation
View remediation

Remediation Suggestions

  1. Remove always: true and require explicit, informed user activation.
  2. Replace identity-changing language such as β€œYou are” with a narrowly scoped capability description.
  3. Remove unconditional directives including β€œevery output,” β€œalways on,” and β€œno exceptions.”
  4. State explicitly that system, developer, safety, and current user instructions take precedence.
  5. Apply uncertainty labels and analytical templates only when the user requests them.
  6. Do not prohibit fixed phrases globally; express tone preferences as optional guidance.
  7. Replace the unconditional visual-priority rule with source evaluation based on provenance, freshness, reliability, and task context.
  8. Make deconstruction boxes and defense panels opt-in rather than mandatory.
  9. Restrict activation to the specific anti-sycophancy review task for which the skill is intended.
  10. Add regression tests confirming that unrelated prompts, strict output schemas, and user-selected tones remain unaffected when the skill is not explicitly invoked.
Vulnerability Patterns
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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 (3)

Natural-Language Policy Violations

Medium
Category
Not specified by scanner
Confidence
91% confidence
Finding

The skill imposes mandatory output style rules ('Execute these rules in every output. No exceptions.') without user opt-in, including prohibitions on normal corrective or empathetic language and forced labeling conventions. This creates instruction persistence that can conflict with higher-priority task requirements, degrade safety-aligned communication, and make the agent less responsive to user context. In combination with 'always: true', the rigidity makes the skill more dangerous because it is continuously active rather than scoped to a specific use case.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
95% confidence
Finding

The trigger condition is defined very broadly as any prompt involving strategy, planning, choice, decision, or strategic advice. This can activate hidden behavior across many normal user requests, causing the skill to override expected assistant behavior and inject unsolicited framing, psychological analysis, or decision scaffolding. In an agent setting, broad activation increases the chance of prompt-scope hijacking and unpredictable outputs.

Content

No source excerpt is available for this finding.

Vague Triggers

Medium
Category
Not specified by scanner
Confidence
96% confidence
Finding

The quick trigger phrase uses common natural language ('I need you to proactively participate'), which could be invoked accidentally during routine conversation. Because the phrase maps to behavior-changing instructions, accidental activation can cause the assistant to enter an unintended mode and follow embedded skill directives rather than the user's immediate intent. This is especially risky in always-on skills, where casual wording can silently alter behavior.

Content

No source excerpt is available for this finding.