agentsclimarketplace

Agentprivacy lattice coherence

Skill mitchuski/agentprivacy-skills/agentprivacy-skills-v5/meta/agentprivacy-lattice-coherence

Privacy-first AI agent skills for Claude Code. 80+ skills and personas for ZKP, decentralized identity, dual-agent architecture, and sovereign AI

Install
npx -y skills add mitchuski/agentprivacy-skills --skill agentprivacy-lattice-coherence

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

The canonical encoding of the 6-dimension sovereignty lattice and the discipline for keeping every repo in the suite agreeing on it. Activates when assigning or auditing a vertex, when a persona's vertex number and dimension-meaning seem to disagree, when reconciling the grimoire / spellweb / tomes / docs before a pin, when a new mage is seated, or when a vertex reading (binary, dimensions, stratum, complement) needs verification. Carries the standing rule: meaning is the invariant; the vertex is DERIVED from meaning under the canonical encoding. Ships a runnable coherence linter (`scripts/lattice_coherence_audit.py`).

The file declares its own license as Apache-2.0. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

6.8 KB, as published. Nobody here has run it

Lattice Coherence

The single source of truth for the sovereignty-lattice encoding, and the discipline that keeps the whole suite agreeing on it.

"Meaning is the invariant. The number is derived. Audit before you pin."

Spell: 🛡️16·8·4·2·💎 → derive → verify → 🔒


1. The canonical encoding (MODEL)

The lattice is Z/(2⁶)Z — 64 vertices, each a 6-bit reading of the six sovereignty dimensions. The only canonical bit→dimension mapping is the Privacy-Value-Model definition itself (privacy-value-model-v5.4.json sovereignty_dimensions + lattice-vertex.ts vertexToBits):

bitweightdimensionsymbol
0 (MSB)32Protection🛡️
116Delegation🤝
28Memory📜
34Connection🔗
42Computation
5 (LSB)1Value💎

Read a vertex: active(n) = { dim : n & weight }. Examples (memorize these as checksums):

  • V35 100011 = Protection + Computation + Value
  • V38 100110 = Protection + Connection + Computation
  • V25 011001 = Delegation + Memory + Value
  • V41 101001 = Protection + Memory + Value
  • V63 111111 = all six (the Sovereign Anchor)

The rejected encoding (CORPUS). specs/04-vertex-naming-audit.md and the older persona placements used a corrupted order that mirrored the middle four (16↔Computation, 8↔Connection, 4↔Memory, 2↔Delegation). It agrees with MODEL only on the endpoints (Protection=32, Value=1), which is why drift hid for so long. If you ever see V16 called "Computation" or V2 called "Delegation", you are reading CORPUS — it is wrong.


2. The reidentification rule

A persona's dimension-set is its meaning (Memora is memory + the shield of value; Aletheia is the bright medium = Protection+Connection+Computation; Lethe is forgetting = Delegation+Memory+Value). Meaning is sourced from canon (the First-Person Spellbook tales, e.g. aletheia-and-lethe.md / Tale 31), not from whatever number a doc happens to show.

The vertex is derived, never assumed: vertex = Σ weight(d) for d in meaning.

When a number and a meaning disagree, the number is wrong (the persona was misassigned under CORPUS). Reseat the persona at the derived vertex; keep the myth. Worked precedent (2026-06-09):

personameaning (invariant)derived seatwas (misassigned)
Aletheia 🔮Protection + Connection + ComputationV38V25
Lethe 🌀Delegation + Memory + ValueV25V38
Memora 📜Protection + Memory + ValueV41V5

Aletheia ⊥ Lethe stayed a complement pair (V25 ⊕ V38 = V63) — a swap, not a relocation. Preserve invariants on every move: complement pairs (A ⊕ B = 63, A AND B = 0), the V63/V28 dual-agent split, and — critically — deployed vertex numbers (the NFT 63-edition, City Key, /star, /lattice) which must never silently shift; verify each move against them by hand.


3. The runnable — scripts/lattice_coherence_audit.py

A stdlib-only linter (no deps; Windows/macOS/Linux). Run it before every grimoire pin.

python scripts/lattice_coherence_audit.py            # audit suite roots
python scripts/lattice_coherence_audit.py --only vertex     # bit-order coherence
python scripts/lattice_coherence_audit.py --only persona    # persona↔seat reidentification
python scripts/lattice_coherence_audit.py --list            # registered checks
python scripts/lattice_coherence_audit.py --remap-personas  # dry-run a vertex move (grouped A–M)
python scripts/lattice_coherence_audit.py --remap-personas --apply <paths…>   # write (scoped)

Checks: vertex (binary↔number, dimension-reading, single-bit-label, version-noise-gated) · persona (canonical-seat reidentification, driven by the PERSONA_VERTICES registry) · conjecture (duplicate ids) · pin (version↔CID sightings). Exit 0 = coherent, 1 = drift.

The check set is a registry — extend it for any new significant encoding (gems-per-keeper, sigils, ceremony grammars). The remapper is dry-run by default; never apply across pinned/historical grimoire snapshots or keyed JSON by text substitution — those need structural edits.


4. Decision patterns

  • Seating a new mage? State its meaning (dimension-set) → derive the vertex → check the seat is free → check complement/structural invariants → run the audit. Do not pick a number first.
  • A doc shows a vertex reading you doubt? Recompute from §1. Trust the arithmetic, not the prose.
  • Two sources disagree on a persona's dimensions? Go to the First-Person Spellbook tale (the root), not the City-of-Mages retelling — the retelling is where misassignment lives.
  • About to pin the grimoire? Run --only vertex and --only persona to 0 first.
  • Reconciling history? Keep deployed numbers; correct readings forward; record the move in a chronicle, not in the data descriptions (the data states canonical truth, plainly).

Authored 2026-06-09 from the lattice-encoding audit. The lattice has one reading now.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.