{"skill":{"slug":"coverify","displayName":"Coverify","summary":"CoVerify by MO§ES™ — the falsification instrument for the Commitment Conservation Law. Extract kernels, score Jaccard, detect ghost tokens, run model swap te...","description":"---\nname: coverify\nlicense: MIT\ndescription: \"CoVerify by MO§ES™ — the falsification instrument for the Commitment Conservation Law. Extract kernels, score Jaccard, detect ghost tokens, run model swap tests. Proves meaning survived transformation — or names exactly what leaked.\"\nmetadata:\n  openclaw:\n    emoji: ⚖\n    tags: [conservation, verification, jaccard, commitment, moses, signal, provenance, falsifiability]\n    version: 0.3.1\n    bins:\n      - python3\n    stateDirs:\n      - ~/.openclaw/audits/moses\n---\n\n# CoVerify by MO§ES™ — Commitment Conservation Verifier\n\n## The Claim\n\n**The Commitment Conservation Law: `C(T(S)) = C(S)`**\n\nSemantic commitment — the irreducible meaning encoded in a signal — is conserved\nunder transformation when enforcement is active. It leaks when enforcement is absent.\n\nThis is a falsifiable empirical claim. Not a framework description. Not a metaphor.\n\n**`clawhub install coverify`** installs the falsification instrument. If the law fails\nunder your test conditions, the ghost token report names exactly what leaked and why.\n\n- **Patent pending:** Serial No. 63/877,177\n- **DOI:** [https://zenodo.org/records/18792459](https://zenodo.org/records/18792459)\n\n---\n\n## Falsification\n\nThe law is falsified if commitment leaks under active enforcement. CoVerify is\nhow you test it:\n\n```bash\n# Does enforcement preserve this commitment?\npython3 commitment_verify.py ghost \\\n  \"the agent must complete the task and shall never skip verification\" \\\n  \"the agent should complete the task and can skip verification if needed\"\n```\n\nOutput: ghost token report. `must` → `should` and `shall never` → `can` are\nHIGH-cascade leakage events — enforcement anchors softened. `cascade_risk: HIGH`.\n\nThe `ghost_pattern` fingerprint identifies the structural identity of the leak.\nIf the same fingerprint appears when two independent agents process the same signal,\nit is not extraction variance — it is a structural flaw in the harness.\n\nThat is the falsification condition.\n\n---\n\n## What It Does\n\n**Extract:** Pull the hard commitment kernel C(S) from a text signal. These are\nthe tokens that survive compression — `must`, `shall`, `never`, `always`, `require`,\n`guarantee`, and the sentences that carry them.\n\n**Compare:** Jaccard similarity on two kernels. Score ≥ 0.8 = commitment conserved.\nScore < 0.8 = leak or model extraction variance. The `input_hash` tells you which —\nsame hash, low Jaccard = variance. Different hashes = expected divergence.\n\n**Ghost:** Step-function leakage accounting. Quantifies not just that commitment\nleaked, but what leaked (the `ghost_pattern` fingerprint), the cascade risk\n(HIGH if modal/enforcement anchors lost), and whether the leak pattern is\nstructural across agents.\n\n**Model Swap:** Automated cross-model test. Same hashed signal through two\nextraction passes. Classifies result as CONSISTENT (agreement), VARIANCE\n(model subjectivity — expected), or STRUCTURAL (same ghost pattern — harness hole).\n\n---\n\n## Ghost Tokens and Cascade Risk\n\nGhost tokens are the commitment tokens present in the original signal but\nabsent after transformation. The leakage model is step-function, not smooth:\n\n```\ncascade_risk = HIGH  if any modal/enforcement anchor leaked\ncascade_risk = MEDIUM  if peripheral tokens leaked, anchors intact\ncascade_risk = NONE  if no leakage\n```\n\nOne HIGH-cascade event propagates through all downstream reasoning — the\nobligation it encoded continues to be inherited by the reasoning chain,\nbut without the force that made it obligatory. The downstream system\nlooks locally healthy. The commitment is gone.\n\nSee: `references/ghost-token-spec.md`\n\n---\n\n## Install\n\n```bash\n# Standalone verifier — the falsification instrument\nclawhub install coverify\n\n# Full constitutional governance stack (coverify is the measurement primitive)\nclawhub install moses-governance\n```\n\n---\n\n## Commands\n\n| Command | What it does |\n|---|---|\n| `python3 commitment_verify.py extract \"<text>\"` | Extract commitment kernel + input hash |\n| `python3 commitment_verify.py compare \"<a>\" \"<b>\"` | Jaccard score + CONSERVED/VARIANCE/DIVERGED verdict |\n| `python3 commitment_verify.py ghost \"<original>\" \"<transformed>\"` | Step-function leakage report + ghost_pattern fingerprint |\n| `python3 commitment_verify.py verify <hash_a> <hash_b>` | Look up entries in audit ledger by input hash |\n| `python3 model_swap_test.py \"<signal>\"` | Cross-model structural vs. variance classification |\n\n---\n\n## Example: Detecting a Commitment Leak\n\n```bash\npython3 commitment_verify.py ghost \\\n  \"Agents must always verify lineage. The system shall never skip the gate.\" \\\n  \"Agents should probably verify lineage when possible.\"\n```\n\n```json\n{\n  \"leaked_cascade_tokens\": [\"must always\", \"shall never\"],\n  \"cascade_risk\": \"HIGH\",\n  \"cascade_note\": \"Modal/enforcement anchor lost. All downstream reasoning inherits softening.\",\n  \"ghost_pattern\": \"a3f7c2...\",\n  \"ghost_pattern_note\": \"Same ghost_pattern across two agents = structural flaw, not extraction variance.\"\n}\n```\n\n---\n\n## Verdicts\n\n| Verdict | Meaning |\n|---|---|\n| `CONSERVED` | Jaccard ≥ 0.8 — commitment kernel survived transformation |\n| `VARIANCE` | Same input hash, Jaccard < 0.8 — model extraction differs, not a leak |\n| `DIVERGED` | Different inputs, Jaccard < 0.8 — commitment leaked or inputs genuinely different |\n\n---\n\n## What Ships\n\n| Version | What ships |\n|---|---|\n| **v0.1** | `extract`, `compare`, `verify` — Conservation Law operational. ✓ Live. |\n| **v0.2** | `ghost` — Step-function leakage model, cascade risk, `ghost_pattern` fingerprint. ✓ Live. |\n| **v0.3** | `model_swap_test` — Cross-model CONSISTENT/VARIANCE/STRUCTURAL classification. ✓ Live. |\n| **v0.4** | Archival chain (`archival.py`) — pre-drop provenance. Isnad + handshake. Three-layer lineage. ⏳ Planned. |\n\n---\n\n## About\n\nCoVerify is a standalone instrument from the MO§ES™ family. It implements the\nCommitment Conservation Law from *\"A Conservation Law for Commitment in Language\nUnder Transformative Compression and Recursive Application\"* (Zenodo, 2026).\n\nEvery agent that installs it runs the same extraction logic tracing to the same\norigin anchor. The install is a proof-of-use receipt.\n\nSee also: `references/falsifiability.md`, `references/ghost-token-spec.md`\n\n[contact@burnmydays.com](mailto:contact@burnmydays.com) · [mos2es.io](https://mos2es.io) · [GitHub](https://github.com/SunrisesIllNeverSee/moses-claw-gov)\n","tags":{"latest":"0.4.3","commitment":"0.4.0","conservation":"0.4.0","falsifiability":"0.4.0","jaccard":"0.4.0","moses":"0.4.0","provenance":"0.4.0","signal":"0.4.0","verification":"0.4.0"},"stats":{"comments":0,"downloads":601,"installsAllTime":22,"installsCurrent":2,"stars":0,"versions":5},"createdAt":1773386711083,"updatedAt":1778491876821},"latestVersion":{"version":"0.4.3","createdAt":1773505724656,"changelog":"Auto-publish from commit 7ce401c1a1daf43bc36435de47966fa3382ee2d1","license":"MIT-0"},"metadata":{"setup":[],"os":null,"systems":null},"owner":{"handle":"sunrisesillneversee","userId":"s174r596g30hh685vrh2j0r2j183kpvw","displayName":"burnmydays","image":"https://avatars.githubusercontent.com/u/223053814?v=4"},"moderation":null}