# How the Anti-Recursion Protocol Handles Recursive Questioning in Council of High Intelligence

> Discover how the anti-recursion protocol in 0xNyk/council-of-high-intelligence prevents infinite loops with depth limits, blocks re-asked questions, and activates the hemlock rule for concise member positions.

- Repository: [nyk/council-of-high-intelligence](https://github.com/0xNyk/council-of-high-intelligence)
- Tags: internals
- Published: 2026-07-03

---

**The anti-recursion protocol prevents infinite deliberation loops by enforcing a three-level questioning depth limit, blocking re-asked questions, and triggering the hemlock rule that forces members to provide concise 50-word positions when recursion is detected.**

The Council of High Intelligence (CoHI) implements strict anti-recursion measures to keep deliberations efficient and prevent members from stalling discussions by repeatedly asking already-answered questions. According to the 0xNyk/council-of-high-intelligence source code, this protocol combines depth limits, transcript scanning, and the hemlock rule to guarantee bounded conversations that cannot spiral into endless recursion.

## Architecture of the Anti-Recursion Protocol

The protocol operates across three distinct layers within the codebase, each responsible for detecting and interrupting recursive patterns before they compromise the deliberation.

### Coordinator Verification Step

In [`SKILL.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/SKILL.md) at line 427, the coordinator executes a `[VERIFY] Anti-recursion` check during the post-round enforcement phase. When the coordinator detects that Socrates—the designated question-asking member—has restated a question that received a direct answer in a previous round, it immediately applies the **hemlock rule**. The flagged member must replace their recursive query with a concise position statement of 50 words or fewer. Additionally, if any member simply repeats their Round 1 answer without engaging new challenges, the coordinator returns the output for revision.

### Socrates Grounding Protocol

The [`agents/council-socrates.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/agents/council-socrates.md) file (lines 24-30) defines the **ANTI-RECURSION** grounding protocol that constrains Socrates specifically. This implementation:
- Limits questioning depth to three levels maximum
- Explicitly disallows re-asking questions that have already been answered
- Enforces the hemlock rule upon recursion detection, forcing Socrates to issue a 50-word strongest position

### README Enforcement Overview

The high-level enforcement mechanism is documented in [`README.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/README.md) at line 241, which summarizes the anti-recursion guard as part of the **Post-Round Enforcement** system. This overview establishes the hard boundaries that cap Socrates' questioning privileges and cut off any member pair that exceeds two back-and-forth messages without resolution.

## Detection and Enforcement Mechanics

The anti-recursion protocol executes through a deterministic four-step process that runs after each round completes.

### Step 1: Transcript Scanning

During the post-round enforcement scan, the coordinator analyzes the complete transcript to identify recursive patterns. The system flags Socrates immediately if he asks a question that mirrors an already-answered query from a previous round or the current session.

### Step 2: Hemlock Rule Application

Upon detecting recursion, the coordinator triggers the **hemlock rule**. The flagged member must immediately abandon their recursive line of questioning and instead provide a short, decisive position statement limited to 50 words or fewer. This constraint forces the discussion forward rather than allowing it to loop back on previously established points.

### Step 3: Restatement Guard

The protocol includes a **restatement guard** that identifies when members attempt to resubmit their Round 1 answers without addressing challenges raised by other council members. The coordinator rejects these submissions and sends them back for revision, requiring substantive engagement with counter-arguments rather than repetition.

### Step 4: Message-Pair Limit

To prevent dialogue ping-pong between two members, the coordinator enforces a strict **message-pair limit**. When any two members exchange more than two messages without reaching convergence, the coordinator terminates that specific exchange immediately. This prevents pairwise conversations from consuming the council's fixed round budget.

These checks execute as Step 4 in [`SKILL.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/SKILL.md), functioning as a hard bound that works alongside the fixed round budget—**full** mode allows 3 rounds, while **quick** mode allows 2 rounds—to guarantee deliberations finish without infinite recursion.

## Source Code Implementation

Three primary files define the anti-recursion protocol's behavior and enforcement boundaries.

| File | Location | Purpose |
|------|----------|---------|
| [`SKILL.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/SKILL.md) | Line 427 | Contains the authoritative coordinator steps including the `[VERIFY] Anti-recursion` rule, hemlock rule enforcement, and message-pair limits |
| [`agents/council-socrates.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/agents/council-socrates.md) | Lines 24-30 | Defines Socrates' specific grounding constraints, explicitly listing the anti-recursion depth limit and hemlock requirements |
| [`README.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/README.md) | Line 241 | Provides the high-level architectural overview of enforcement mechanisms visible to end users |

## Practical Implementation Examples

The following pseudo-code illustrates the logical flow the coordinator follows during the post-round enforcement scan:

```python

# Coordinator logic during Step 4: Post-Round Enforcement

for member, msg in round_transcripts:
    if member == "Socrates" and is_recursive_question(msg):
        # Trigger hemlock rule on recursion detection

        send_back(member,
                  "HEMLOCK: State your strongest position in ≤50 words.")
        continue
    
    if repeats_round1(member, msg):
        # Block simple restatements without engagement

        send_back(member,
                  "Please address the challenges raised; do not simply restate Round 1.")
        continue
    
    if exchange_exceeds_pair_limit(member, other, msg):
        # Cut off diverging pairwise dialogues

        terminate_pair(member, other)

```

When the hemlock rule fires, Socrates must output a constrained position statement formatted as follows:

```yaml
---
Hemlock Position: |
  The core issue is that the current deployment strategy assumes
  linear scaling of performance, which contradicts observed non‑linear
  load patterns. We must adopt a tiered scaling model that accounts for
  variable demand, otherwise we risk severe over‑provisioning.

```

Conversely, when a member attempts to restate their Round 1 position without engaging challenges, the coordinator returns this error:

```yaml
---
Error: "Your reply does not engage any of the challenges raised by other members.
Please incorporate at least one new argument or rebuttal."

```

## Summary

- **The anti-recursion protocol** enforces bounded deliberations through three specific mechanisms: depth limits, the hemlock rule, and message-pair caps.
- **The hemlock rule** triggers when Socrates asks already-answered questions, forcing a 50-word maximum response that moves the conversation forward.
- **Source implementation** spans [`SKILL.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/SKILL.md) (coordinator logic), [`agents/council-socrates.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/agents/council-socrates.md) (agent constraints), and [`README.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/README.md) (architectural overview).
- **Hard boundaries** include a three-level questioning depth, two-message limits between member pairs, and fixed round budgets (2-3 rounds total).
- **Post-round enforcement** scans occur after every round to catch and correct recursive patterns before they propagate.

## Frequently Asked Questions

### What triggers the hemlock rule in the Council of High Intelligence?

The hemlock rule triggers when the coordinator detects that Socrates has asked a question that has already received a direct answer in a previous round or when the questioning exceeds the three-level depth limit. According to [`SKILL.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/SKILL.md) line 427 and [`agents/council-socrates.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/agents/council-socrates.md) lines 24-30, this rule forces the flagged member to immediately provide a concise position statement of 50 words or fewer, preventing the discussion from cycling back to previously resolved points.

### How does the message-pair limit prevent infinite loops?

The message-pair limit acts as a circuit breaker when two members exchange more than two messages without reaching agreement. As implemented in the coordinator verification step, this limit prevents pairwise dialogues from devolving into endless ping-pong arguments that consume the council's fixed round budget, ensuring that debates either converge or terminate within strict bounds.

### What happens if a member simply repeats their Round 1 answer?

The coordinator's restatement guard detects when members resubmit their initial positions without addressing challenges raised by other council members. Rather than allowing this recursive restatement, the system bounces the output back for revision with a specific error message requiring the member to incorporate at least one new argument or rebuttal, as defined in the post-round enforcement logic.

### Can the anti-recursion protocol be disabled or configured?

Based on the source code in [`SKILL.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/SKILL.md) and [`agents/council-socrates.md`](https://github.com/0xNyk/council-of-high-intelligence/blob/main/agents/council-socrates.md), the anti-recursion protocol represents a hard-coded constraint system rather than a configurable feature. The three-level depth limit, 50-word hemlock constraint, and two-message pair limit are architectural constants designed to guarantee that all deliberations—whether in full mode (3 rounds) or quick mode (2 rounds)—terminate predictably without infinite recursion.