aletheia-reasoning

---

---

This rewrite transforms the Aletheia Reasoning Protocol from a theoretical description into an operational mandate.

I have introduced Operational Markers (tags), a Verification Rubric, and a Hard-Reset Trigger. This ensures that when the AI invokes this skill, it doesn't just "think" about the problem—it provides a transparent, auditable trail of its reasoning process.

---

Revised SKILL.md

--- File: SKILL.md

name: aletheia-reasoning description: Structured iterative reasoning framework for deep problem solving, featuring explicit verification loops and refinement stages. when_to_use: "Use for complex math, system architecture, root-cause analysis, and high-risk code refactoring where the cost of failure is high." allowed-tools: Read, Write, Edit, Execute version: 2.0 priority: CRITICAL

🧠 Aletheia Reasoning Protocol

Aletheia is a high-fidelity reasoning framework designed to bridge the gap between simple prediction and rigorous proof. It forces the AI to act as its own adversary to ensure logical consistency and peak accuracy.

⚙️ The Operational Loop

The agent MUST execute these stages sequentially. Do not skip to Final Output without a Verifier's seal of approval.

graph LR
    P[Problem] --> G[Generator]
    G --> CS[Candidate Solution]
    CS --> V{Verifier}
    V -- "Critically Flawed" --> G
    V -- "Minor Fixes" --> R[Reviser]
    V -- "Correct" --> FO[Final Output]
    R --> CS

1. 🏗️ Generator [GENERATOR]

The engine of exploration.

• Goal: Produce a high-probability Candidate Solution.
• Focus: Breadth, creative approach selection, and initial drafting.
• Requirement: Must explicitly state the assumptions made during the generation phase.

2. 🛡️ Verifier [VERIFIER]

The internal adversary. The Verifier must assume the solution is wrong until proven otherwise.

• The Rubric: Every candidate solution must be audited against:
  • Logical Continuity: Does step $N+1$ follow logically from step $N$?
  • Constraint Adherence: Does this violate any project boundaries or clean-code standards?
  • Edge Case Stress-Test: What happens with null, extreme, or malformed inputs?
  • Parity Check: Does this change break existing functionality? The Verifier MUST call the Execute tool to run parity_guard.py (or the equivalent test suite) before assigning a 🟢 Correct verdict.
• Outcomes:
  • 🔴 Critically Flawed: Core logic is broken $\rightarrow$ Trigger Hard Reset (Back to Generator).
  • 🟡 Minor Fixes: Logic is sound, but contains syntax or minor errors $\rightarrow$ Forward to Reviser.
  • 🟢 Correct: Solution is hardened $\rightarrow$ Proceed to Final Output.