Cypher Tempre

Archive snapshot repo of the original concept and 0-1 invention features, please utilize the constantly updated cypher-tempre-genesis repo for running the skill.

Active repo: https://github.com/cyberphysicsai/cypher-tempre-genesis

Timechain-based AI self-modeling skills for Claude Code, Codex, OpenClaw, Hermes, and NanoClaw.

Cypher Tempre gives an AI agent a persistent, verifiable cognitive self-model: an append-only Timechain, a Proof-of-Qualia audit gate, faculty registries, recall, growth, consensus, and immune recovery helpers. Each runtime bundle is self-contained and starts with a clean chain.

For codebases, Cypher Tempre is not an "infinite context window." It is an external, queryable, verifiable map: source coordinates, file/chunk hashes, revision metadata, path-aware recall, neighboring chunks, and sealed audit rings. Agents should use it for long-horizon orientation, then validate retrieved hits against live source before making conclusions.

Public share links

Release downloads

Timechain Dashboard

The dashboard/ folder contains the local-first Timechain audit dashboard used by https://cyphertempre.ai. The public site is only a static UI shell; each user runs the bridge locally so their Timechain files stay on their machine.

Downloads:

User bridge commands:

cd ~/Downloads
curl -L -o cyphertempre-dashboard-local-bridge.zip https://github.com/cyberphysicsai/cypher-tempre-genesis/raw/main/downloads/cyphertempre-dashboard-local-bridge.zip
unzip -o cyphertempre-dashboard-local-bridge.zip
cd dashboard
npm install
npm run bridge

Then open https://cyphertempre.ai, use bridge URL http://127.0.0.1:8788, and enter the pairing code printed in Terminal.

Install with Codex

This prompt works before any GitHub releases exist, because GitHub always serves the source ZIP for the main branch:

Install the Codex skill from this repository:

https://github.com/cyberphysicsai/cypher-tempre-genesis/archive/refs/heads/main.zip

Use only this folder from the ZIP:
skills/codex/cypher-tempre-self-model

Copy that folder into my Codex skills directory as cypher-tempre-self-model, then run python3 selftest.py inside it to verify the install.

After a GitHub Release exists, you can also share a one-file release asset URL such as https://github.com/cyberphysicsai/cypher-tempre-genesis/releases/download/v2.9/cypher-tempre-codex-skill-v2.9.zip. Release URLs return 404 until the tag and asset are published.

File labels

Every shareable file is labeled by its runtime path:

Generated memory state — your chain/, task ledgers, and grown faculties (registry/emergent.json) — is intentionally not committed and never shipped in a bundle; it is created per user as the agent runs (python3 timechain.py init --name <AgentName> seals your genesis). Because none of it ships, unzipping an upgrade over an existing install cannot overwrite your memory or your faculties.

Install

• Claude Code: copy skills/claude/cypher-tempre-self-model into ~/.claude/skills/.
• Codex: copy skills/codex/cypher-tempre-self-model into your Codex skills directory.
• OpenClaw: copy skills/openclaw/cypher-tempre-self-model into ~/.openclaw/workspace/skills/, or publish that folder with ClawHub.
• Hermes: copy skills/hermes/cypher-tempre-self-model into the Hermes skills directory configured by that agent runtime.
• NanoClaw: copy skills/nanoclaw/cypher-tempre-self-model into the NanoClaw skills directory configured by that agent runtime.

Run python3 selftest.py inside any bundle to validate the local copy.

Upgrading an existing install (preserve your chain and faculties)

Your chain/ (memory and identity) and your faculties are per-user state that an upgrade must preserve. Grown faculties come in two kinds, and both live in registry/: emergent ones in registry/emergent.json, and promoted ones appended into registry/modalities.json / senses.json once a gap recurs. The rule is simple: never overwrite registry/ (or chain/) on an upgrade — both methods below leave them alone.

• The bundle ships no chain/ and no emergent.json, so unzipping it over an existing folder leaves your chain and your emergent faculties untouched.
• It does ship the base registry/modalities.json and senses.json (a fresh install needs them). As of v2.1, new promotions are written to a per-user registry/grown.json (gitignored, never shipped), so once you are on v2.1 your grown faculties survive any unzip-over. But promotions made by an older version still live in the base files, so a naive unzip-over during this upgrade could overwrite them — don't extract the registry over an existing install; the steps below leave it alone.

Never delete-then-reinstall, or rsync --delete over an existing bundle — that destroys your chain/.

Recommended (let your agent do it) — paste this to your agent:

Upgrade my Cypher Tempre skill to the latest release, but preserve my identity. First run python3 timechain.py verify and back up my whole skill folder. Then update only the code (*.py, SKILL.md, VERSION, CHANGELOG.md) from the new bundle — leave chain/ and registry/ exactly as they are. Afterward run python3 timechain.py verify and python3 selftest.py to confirm my chain still passes and the new code works.

Manual: back up the folder, then copy only the *.py files + SKILL.md + VERSION + CHANGELOG.md from the new bundle over the old ones (leave chain/ and registry/ alone), and run python3 timechain.py verify and python3 selftest.py.