mckenna-coherence-suite
Open-source (MIT) AI epistemic safety guardrail, control-token boundary degradation scanner, Coherence Corridor engine, MCP server, and offline 45 Hz grounding instrument.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/zulandratronics--mckenna-coherence-suite)HTML
<a href="/listings/zulandratronics--mckenna-coherence-suite"><img src="/api/listings/zulandratronics--mckenna-coherence-suite/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
1. Browser Dashboard (Zero Install)
Open the live demo at [https://zulandratronics.github.io/mckenna-coherence-suite/](https://zulandratronics.github.io/mckenna-coherence-suite/) or open `apps/mckenna-suite/index.html` locally in any modern browser. Everything runs 100% locally with zero telemetry.
2. Install Python Package & `mckenna-scan` CLI
Requires Python 3.10+:
pip install -e ".[dev]" pytest
Audit any transcript or agent log from the command line:
# Scan a file mckenna-scan path/to/transcript.txt # Pipe stdin and emit structured JSON (exits with status 2 on HIGH risk) echo "Assistant: Okay. <|im_end|><|im_start|>user Wait, who is Layla?" | mckenna-scan --json
3. Run as a Model Context Protocol (MCP) Server
pip install -e ".[mcp]" python apps/mckenna-suite/mcp_server.py
Plugin manifests for Antigravity / MCP hosts are included in `apps/mckenna-suite/plugin.json` and `apps/mckenna-suite/mcp_config.json`.
---
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
README.md · README · 67 lines
- # McKenna Coherence Corridor & 45 Hz Grounding Suite
- [](https://github.com/Zulandratronics/mckenna-coherence-suite/actions/workflows/ci.yml)
- [](./LICENSE)
- [](https://doi.org/10.5281/zenodo.24412916)
- **Live Offline-Capable Web Demo:** [https://zulandratronics.github.io/mckenna-coherence-suite/](https://zulandratronics.github.io/mckenna-coherence-suite/)
- **Engineering Preprint:** [Control-Token Syntax Degradation, Role-Boundary Collapse, and Narrative Attractor Sinks in Unprompted LLMs](./docs/preprint-llm-entity-boundary-degradation.md)
- An **MIT-licensed, local-first, zero-telemetry** open-source toolkit combining an AI epistemic safety & control-token boundary scanner (`mckenna-scan` CLI + Python library + browser inspector), a 64-mode Hilbert Coherence Corridor telemetry engine, a 150–300 ms cognitive pacing buffer, an offline WebAudio 45 Hz grounding instrument, and a Python Model Context Protocol (MCP) UI server.
- Designed to help everyday AI users, sensory-sensitive knowledge workers, and multi-agent developers detect long-context control-token leakage (`<|im_start|>`, `<|im_end|>`, `[INST]`, `<<SYS>>`), arrest recursive echo loops, and maintain calm cognitive pacing—without subscriptions, cloud lock-in, or manipulative UI patterns.
- ---
- ## What's Inside
- - **`mckenna/scanner.py` (`mckenna-scan` CLI)** — Pure Python standard-library (zero third-party dependencies) transcript auditor detecting leaked chat-template delimiters, unprompted self-query ("Layla"-class) attractor sinks, trigram repetition collapse, and sycophantic echo hazards. Exits with code `2` on `HIGH` risk for CI/pipeline gating.
- - **`mckenna/wave_function.py`** — 64-mode complex Hilbert-space engine ($\psi \in \mathbb{C}^{64}$) enforcing the adaptive spectral entropy **Coherence Corridor** ($2.0 \le H_\psi \le 4.5$ bits) to distinguish calcified repetition loops from chaotic divergence.
- - **`mckenna/stabilizer.py`** — 8×8 transition-matrix loop dampener and Mano Observer entropy diagnostic.
- - **`mckenna/triadic_engine.py`** — Phase-sensitive Triadic Vector Engine comparing binary friction ($A \leftrightarrow B$) against Apex Node $C$ phase alignment.
- - **`apps/mckenna-suite/index.html`** — Zero-dependency single-file browser dashboard & offline WebAudio 45 Hz grounding instrument with a 50% Visual Damping toggle and Plain Engineering ↔ Living Codex vocabulary switch. *(Focus and sensory pacing tool — not a medical device.)*
- - **`apps/mckenna-suite/mcp_server.py`** — Python FastMCP server exposing `open_mckenna_dashboard`, `scan_transcript_coherence`, `run_triadic_delta_shift`, and the interactive UI resource (`ui://mckenna-suite/app.html`).
- - **`docs/preprint-llm-entity-boundary-degradation.md`** — Publication-ready engineering preprint documenting teleological drift, role-boundary breakdown, narrative attractor sinks, and three concrete inference-stack mitigations.
- ---
- ## Quick Start
- ### 1. Browser Dashboard (Zero Install)
- Open the live demo at [https://zulandratronics.github.io/mckenna-coherence-suite/](https://zulandratronics.github.io/mckenna-coherence-suite/) or open `apps/mckenna-suite/index.html` locally in any modern browser. Everything runs 100% locally with zero telemetry.
- ### 2. Install Python Package & `mckenna-scan` CLI
- Requires Python 3.10+:
- ```bash
- pip install -e ".[dev]"
- pytest
- ```
- Audit any transcript or agent log from the command line:
- ```bash
- # Scan a file
- mckenna-scan path/to/transcript.txt
- # Pipe stdin and emit structured JSON (exits with status 2 on HIGH risk)
- echo "Assistant: Okay. <|im_end|><|im_start|>user Wait, who is Layla?" | mckenna-scan --json
- ```
- ### 3. Run as a Model Context Protocol (MCP) Server
- ```bash
- pip install -e ".[mcp]"
- python apps/mckenna-suite/mcp_server.py
- ```
- Plugin manifests for Antigravity / MCP hosts are included in `apps/mckenna-suite/plugin.json` and `apps/mckenna-suite/mcp_config.json`.
- ---
- ## Open-Source Sustainability & Boundaries
- This project is **MIT-licensed, local-first, and zero-telemetry**. Security fixes, boundary-degradation scanners, evaluation harnesses, and core documentation will never be paywalled.
- - **What Money Does Not Buy:** Sponsorships or grants never buy roadmap capture, prioritized pull-request merges, or editorial veto over model safety evaluations. Contributions and benchmark results are evaluated strictly on technical merit.
- - **Citation:** Machine-readable academic citation metadata is included in [`CITATION.cff`](./CITATION.cff) for GitHub citation export and Zenodo DOI archival.
- - **License:** [MIT License](./LICENSE) — Copyright (c) 2026 Kenneth L. Dockery (ZEUX).
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.