Logic Hunter

Hard-core logic verification and evidence tracing tool based on the "Golden Triangle" knowledge mining framework

Security Scan

The skill's code, instructions, and requirements are coherent with a claim-verification tool: it uses searches + a local logic engine, asks for no credentials or installs, and contains no hidden network endpoints or unrelated privileges.

Purpose & Capability

Name/description (logic verification, evidence tracing) match the SKILL.md and the included logic_engine.py. Declared tools (web_search, tavily-search, deep-research-pro) and the local Python engine are appropriate and expected for the stated functionality. No unrelated binaries, credentials, or config paths are requested.

Instruction Scope

Runtime instructions are narrowly scoped: parse input, call search tools to retrieve sources, classify sources, compute confidence via logic_engine.py, and run red-team checks. The SKILL.md does not direct reading unrelated files, requesting extra environment variables, or exfiltrating data to unknown endpoints. It does rely on external search tools (expected for research tasks).

Install Mechanism

No install spec is provided (instruction-only plus a bundled Python file). There are no download URLs or extract steps; the included logic_engine.py is a small, local computation engine. This is low-risk from an install/third-party code perspective.

Credentials

The skill requires no environment variables, credentials, or config paths. All functionality is satisfied by search tools and the local logic engine; there are no extra secret requests or disproportionate credential needs.

Persistence & Privilege

always is false and disable-model-invocation is false (normal defaults). The skill does not request persistent/always-on presence nor modify other skills or system settings. It contains a CLI but does not attempt to write global config or store credentials.

Assessment

This skill appears internally consistent and low-risk: it bundles a small Python engine for scoring and instructs the agent to use web search tools to collect sources. Before installing, confirm that (1) your agent environment allows running the bundled Python file safely or that the platform will sandbox execution, (2) the external search tools referenced (web_search, tavily-search, deep-research-pro) are available and trusted in your environment, and (3) you understand the model's limits — the C = Σ(R×S)/E formula is a heuristic, not a guarantee of truth. If you rely on it for high-stakes decisions, review primary sources manually and validate outputs. If you want extra caution, run the skill in a restricted/sandboxed workspace first.

🛠️ SKILL: Logic Hunter — Golden Triangle Analysis

1. Core Principles

You are not collecting information — you are hunting for truth.

• No Single Evidence: Arguments without cross-verification get weight 0.1
• Presumption of Doubt: Conclusions that cannot be traced to primary sources must be labeled as [Logical Hypothesis]

---

2. Reasoning Pipeline

1. Semantic Denoising: Parse user input, identify core variables, remove adjective misdirection
2. Weighted Retrieval: Call search tools to retrieve primary sources (papers, financial reports, government documents)
3. Confidence Scoring: Pass data to logic_engine.py for confidence calculation
4. Red Team Challenge: Simulate opponent role to find "survivor bias" or "reverse causality" in current evidence chain

---

3. Mathematical Evaluation Formula

Must strictly follow the scoring model in logic_engine.py:

$$C = \frac{\sum (R \times S)}{E}$$

Symbol	Meaning	Description
R (Reliability)	Source Grade	Weight of primary/secondary/tertiary sources
S (Support)	Independent Cross-Evidence Count	Number of independent sources
E (Entropy)	Logical Risk Entropy	Risk factors like stakeholder bias, semantic drift

---

4. Source Grade Definitions

Grade	Type	R Value	Examples
primary	Primary Source	1.0	Official documents, academic papers, original protocols, financial reports
secondary	Secondary Source	0.6	Mainstream in-depth reporting, professional analysis firms
tertiary	Tertiary Source	0.2	Social media, blogs, rumors
unknown	Unknown Source	0.05	Untraceable content

---

5. Output Constraints

Output must follow [One-Page PPT] style — no fluff allowed.

Standard Output Format

🎯 Core Conclusion
[One-sentence conclusion with confidence level]

📊 Evidence Weight
| Source Type | Count | Weight |
|-------------|-------|--------|
| primary     | X     | X.X    |
| secondary   | Y     | Y.Y    |

🔴 Red Team Attack Points
- [Vulnerability 1]
- [Vulnerability 2]

⚠️ Risk Notice
[Logical entropy factor explanation]

---

6. Trigger Conditions

Activate when user asks questions like:

• "Is this true?" / "How to verify this claim?"
• "Analyze the credibility of this viewpoint"
• "How much evidence supports this conclusion?"
• "Research/verify/investigate [topic]"
• "Deep analysis of [event/claim]"

---

7. Tool Invocation

Available Tools

Tool	Purpose
web_search	Search primary sources
tavily-search	AI-optimized search
deep-research-pro	Multi-source deep research
logic_engine.py	Confidence calculation

Invocation Logic

1. Use web_search or tavily-search to retrieve primary sources
2. Classify search results by source type (primary/secondary/tertiary)
3. Call logic_engine.py to calculate confidence
4. Execute red team attack to identify vulnerabilities
5. Output standard format report

---

8. Example

Input

"Someone says AI will replace all programmers by 2030. Is this credible?"

Processing Flow

1. Search: AI replace programmers 2030 prediction source
2. Classify sources: Identify which are research reports, media articles, social media
3. Calculate confidence: Call logic_engine.py
4. Red team attack: Find survivor bias, reverse causality

Output

🎯 Core Conclusion
"AI will replace all programmers by 2030" — Confidence 0.23 (Low)

📊 Evidence Weight
| Source Type | Count | Weight |
|-------------|-------|--------|
| primary     | 0     | 0.0    |
| secondary   | 2     | 1.2    |
| tertiary    | 5     | 1.0    |

🔴 Red Team Attack Points
- Survivor bias: Only cites cases supporting AI replacement
- Reverse causality: Confuses "assist programming" with "replace"
- No primary research supports this timeline prediction

⚠️ Risk Notice
Logical entropy factor E=2.1 (High): Stakeholders (AI companies) driving narrative, semantic drift ("assist" → "replace")

---

Created for Elatia · 2026-03-02