scrimshawlife-ctrl · Sep 5, 2026 · Sep 5, 2026
@@ -0,0 +1,22 @@
+{
+  "name": "hyperlex",
+  "displayName": "Hyperlex",
+  "version": "0.4.2",
+  "description": "Memetic emergence analysis: slang detection, lineage matching, virality and hyperstition scoring, receipt-backed forecasts, and settled Brier calibration. Never invents Brier scores; settlement stays a human step.",
+  "author": {
+    "name": "Applied Alchemy Labs / Hermes"
+  },
+  "homepage": "https://github.com/Zero-State-LLC/Hyperlex",
+  "repository": "https://github.com/Zero-State-LLC/Hyperlex",
+  "license": "LicenseRef-Zero-State-Proprietary-1.0",
+  "keywords": [
+    "memetics",
+    "slang",
+    "hyperstition",
+    "virality",
+    "lineage",
+    "calibration",
+    "brier",
+    "mutation"
+  ]
+}
@@ -0,0 +1,29 @@
+---
+name: hyperlex-analyze
+description: Analyze pasted text or a JSON ingest file with the Hyperlex CLI. Prefer offline route.
+when_to_use: User pastes slang/text to analyze, or points at a file for Hyperlex analyze.
+allowed-tools: Bash Read
+arguments: [input]
+---
+
+# /hyperlex-analyze
+
+Analyze text or a file. Prefer `pipeline` / `run` when the user wants a receipt + forecasts. Use `analyze` for analysis-only.
+
+```bash
+: "${HLX:=python3 ${HYPERLEX_SKILL_DIR:-${HERMES_SKILL_DIR:-$HOME/.claude/skills/hyperlex}}/scripts/hyperlex.py}"
+# pasted text
+$HLX analyze "$ARGUMENTS" --route offline
+# full packet (receipt + forecasts)
+$HLX pipeline "$ARGUMENTS" --route offline
+# ingest JSON file
+$HLX analyze --input "$ARGUMENTS" --route offline
+```
+
+## Expected
+
+JSON result. `provenance.brier` is `null`. Lineage / virality / mutation_trace only when the CLI attaches them.
+
+## Fail-closed
+
+Do not invent labels or Brier. If the file is missing, say so. Do not settle.
@@ -0,0 +1,29 @@
+---
+name: hyperlex-demo
+description: Run the Hyperlex offline demo and summarize receipt labels. No API keys. Never invent Brier.
+when_to_use: User asks for a Hyperlex demo, first success path, or a sample receipt.
+allowed-tools: Bash Read
+---
+
+# /hyperlex-demo
+
+Run the offline demo. Summarize labels from the CLI output. Do not invent scores.
+
+```bash
+: "${HLX:=python3 ${HYPERLEX_SKILL_DIR:-${HERMES_SKILL_DIR:-$HOME/.claude/skills/hyperlex}}/scripts/hyperlex.py}"
+$HLX demo
+# repo checkout: python3 scripts/hyperlex.py demo
+# wrapper: bash scripts/claude_hlx.sh demo
+```
+
+## Expected
+
+- Exit `0`, `ok: true`
+- `brier` and `provenance_brier` are `null`
+- Receipt path under the demo out dir or `~/.hyperlex/receipts/`
+- Known atoms often match a lineage family (`rizz` → `brainrot-aura`)
+- Repeat OBSERVED / INFERRED / SPECULATIVE — do not upgrade them
+
+## Fail-closed
+
+If the CLI fails, report the error. Do not fabricate a receipt or Brier. Do not settle.
@@ -0,0 +1,24 @@
+---
+name: hyperlex-pending
+description: List open Hyperlex forecasts from the score log. Start of the calibration loop.
+when_to_use: User asks what forecasts are open, pending, or ready to settle.
+allowed-tools: Bash Read
+---
+
+# /hyperlex-pending
+
+List unsettled forecasts. This is the start of the calibration loop — not settlement.
+
+```bash
+: "${HLX:=python3 ${HYPERLEX_SKILL_DIR:-${HERMES_SKILL_DIR:-$HOME/.claude/skills/hyperlex}}/scripts/hyperlex.py}"
+$HLX pending
+$HLX pending --limit 20
+```
+
+## Expected
+
+JSON list of open forecast ids (default log: `~/.hyperlex/score_log.jsonl`). Empty list is valid.
+
+## Fail-closed
+
+Do not invent forecast ids. Do not compute Brier here. Next step is human: `/hyperlex-settle` after they choose TRUE|FALSE|VOID|CONFLICT.
@@ -0,0 +1,28 @@
+---
+name: hyperlex-scan
+description: LIVE_EMERGENCE_SCAN helper. Pass a query (or comma-separated queries). Offline-first.
+when_to_use: User wants a Hyperlex scan, live-emergence scan, or multi-query monitor run.
+allowed-tools: Bash Read
+arguments: [query]
+---
+
+# /hyperlex-scan
+
+Multi-query LIVE_EMERGENCE_SCAN. Prefer `--route offline` unless the user allows network.
+
+```bash
+: "${HLX:=python3 ${HYPERLEX_SKILL_DIR:-${HERMES_SKILL_DIR:-$HOME/.claude/skills/hyperlex}}/scripts/hyperlex.py}"
+$HLX scan --query "$ARGUMENTS" --route offline --receipt --forecasts --append-log
+# several terms:
+$HLX scan --queries "rizz,locked in" --route offline --receipt --forecasts --append-log
+# bundled config (repo / installed skill):
+$HLX scan --config "${HYPERLEX_SKILL_DIR:-.}/examples/cron/scan-queries.json" --route offline --receipt --forecasts
+```
+
+## Expected
+
+JSON scan summary. Per-query receipts when `--receipt` is set. Forecasts have `brier: null` until settle.
+
+## Fail-closed
+
+Do not register cron. `risk-schedule` is advisory only. Do not invent Brier. Do not auto-settle.
@@ -0,0 +1,29 @@
+---
+name: hyperlex-settle
+description: Coach Hyperlex settlement. Human must supply forecast-id and decision. Never invent outcomes or Brier.
+when_to_use: User wants to settle a forecast or close the calibration loop.
+allowed-tools: Bash Read
+---
+
+# /hyperlex-settle
+
+Settlement is an **explicit human step**. Do not invent TRUE/FALSE/VOID/CONFLICT. Do not invent Brier.
+
+1. `$HLX pending` — show open ids.
+2. Ask the operator for `--forecast-id` and `--decision`.
+3. Only after they choose, run:
+
+```bash
+: "${HLX:=python3 ${HYPERLEX_SKILL_DIR:-${HERMES_SKILL_DIR:-$HOME/.claude/skills/hyperlex}}/scripts/hyperlex.py}"
+$HLX settle --forecast-id <id> --decision TRUE
+# FALSE | VOID | CONFLICT
+$HLX score-series --mean-shift --verify-chain
+```
+
+## Expected
+
+Settle appends the score log. `score-series` is the only place a numeric Brier is real. Empty series → `NOT_COMPUTABLE`.
+
+## Fail-closed
+
+If the operator has not chosen a decision, stop. Coach the flags. Do not auto-settle. Do not guess outcomes.
@@ -0,0 +1,26 @@
+---
+name: hyperlex-wizard
+description: Week-one Hyperlex onboarding via wizard --auto. Offline. Never auto-settles.
+when_to_use: User is new to Hyperlex, asks to get started, or wants the guided operator path.
+allowed-tools: Bash Read
+---
+
+# /hyperlex-wizard
+
+Week-one guided path. Prefer `--auto` in Claude (non-interactive).
+
+```bash
+: "${HLX:=python3 ${HYPERLEX_SKILL_DIR:-${HERMES_SKILL_DIR:-$HOME/.claude/skills/hyperlex}}/scripts/hyperlex.py}"
+$HLX wizard --auto
+$HLX wizard --auto --query "<term>"
+```
+
+## Expected
+
+Steps: `env_intro` → `doctor` → `demo` → `first_pipeline` → `calibration_coach` → `score_series_hint` → `handoff`.
+
+Summarize those steps. Open analysis stays `brier: null`. Show any open forecasts from the wizard / `$HLX pending`.
+
+## Fail-closed
+
+Do not auto-settle. Ask the operator for TRUE|FALSE|VOID, then coach `$HLX settle`. Do not invent Brier.
@@ -0,0 +1,43 @@
+---
+name: hyperlex
+description: >
+  Main Hyperlex contract in Claude Code. Use for slang, memetics, lineage,
+  virality, hyperstition, receipts, forecasts, and settlement coaching.
+  Prefer the CLI. Not for jailbreak / wrap composition.
+when_to_use: >
+  User asks for Hyperlex, slang analysis, memetic emergence, lineage,
+  Brier settlement, or cultural-signal receipts.
+allowed-tools: Bash Read Grep
+---
+
+# Hyperlex (project skill)
+
+Read the repo-root [SKILL.md](../../../SKILL.md) for the full Hermes + Claude
+contract. This file is a thin pointer so Claude Code can invoke `/hyperlex`
+when this checkout is the project. Do not duplicate engine logic here.
+
+## Invoke
+
+```bash
+export HYPERLEX_SKILL_DIR="${HYPERLEX_SKILL_DIR:-$PWD}"
+export HERMES_SKILL_DIR="${HERMES_SKILL_DIR:-$HYPERLEX_SKILL_DIR}"
+export HLX="${HLX:-python3 $HYPERLEX_SKILL_DIR/scripts/hyperlex.py}"
+$HLX check
+$HLX doctor
+$HLX demo
+$HLX wizard --auto
+$HLX pipeline "<term>" --route offline
+```
+
+Installed personal skill:
+
+```bash
+export HLX="python3 $HOME/.claude/skills/hyperlex/scripts/hyperlex.py"
+# or: bash "$HOME/.claude/skills/hyperlex/scripts/claude_hlx.sh"
+```
+
+## Fail-closed
+
+- Never invent a numeric Brier. Open analysis keeps `brier: null`.
+- Never auto-settle. Ask for TRUE|FALSE|VOID|CONFLICT, then run `settle`.
+- Prefer Bash CLI over guessing labels. Repeat OBSERVED / INFERRED / SPECULATIVE.
@@ -38,6 +38,7 @@ build/
 !docs/command-router.v1.json
 !data/backfill/**/*.json
 !data/phylogeny/**/*.json
+!.claude-plugin/plugin.json
 
 # Temp / build output
 tmp/

@@ -2,6 +2,13 @@
 
 ## Unreleased
 
+- **Claude Code host (additive):** `.claude-plugin/plugin.json`, project
+  `CLAUDE.md`, slash helpers (`.claude/skills/` + plugin `commands/`),
+  `install.sh --claude` / `--claude-plugin`, `scripts/claude_hlx.sh`,
+  `docs/claude-skill.md`, `docs/claude-runtime-contract.md`,
+  `references/claude-runtime-contract.md`.
+  `doctor` reports `CLAUDE_OK` / `CLAUDE_MISSING` (missing does not fail).
+  Hermes install paths unchanged.
 - CLI `wizard` + package `hyperlex.wizard`: week-one Hermes guided path
   (`--auto` / interactive); never auto-settles; offline-first; SKILL.md procedure
 - **SIGNAL REPORT parity (Companion adaptation):** `result.v1` extended with optional `provenance.seed`, `analysis.compression_metrics`, `analysis.symbolic_role`, `analysis.propagation_vector`, `analysis.slang_family_tree`, `analysis.signal_report` (schema + package-local copy). Builder: `src/hyperlex/analysis/signal_report.py`. Wired into `detect_memetic_patterns` (attach + seed header). Docs: `docs/superpowers/specs/2026-08-06-signal-report-adaptation.md`. All new fields optional and fail-open. Brier remains null on open analysis.

@@ -1,2 +1,14 @@
 Follow [AGENTS.md](AGENTS.md). That file is the only contributor and agent instruction source in this repository.
 For a non-trivial change, read the active intent.md before editing. If none exists, draft intent.md and wait for human accept.
+
+---
+
+## Claude Code host (additive)
+
+When using Claude Code as an additional host (Hermes remains primary), see
+[`docs/claude-skill.md`](docs/claude-skill.md) and
+[`references/claude-runtime-contract.md`](references/claude-runtime-contract.md).
+Install helpers: `bash install.sh --claude` / `--claude-plugin`.
+Wrapper: `bash scripts/claude_hlx.sh <command>`.
+
+AGENTS.md remains the only contributor/agent instruction source for this org repo.
@@ -27,6 +27,20 @@ $HLX wizard --auto
 $HLX commands
 ```
 
+## Claude Code
+
+```bash
+bash install.sh --claude
+export HYPERLEX_SKILL_DIR="${HOME}/.claude/skills/hyperlex"
+export HLX="python3 $HYPERLEX_SKILL_DIR/scripts/hyperlex.py"
+$HLX demo
+$HLX wizard --auto
+```
+
+Personal skill path: `~/.claude/skills/hyperlex`. Plugin path:
+`~/.claude/plugins/hyperlex` (`--claude-plugin`). Project helpers live under
+`.claude/skills/`. Settlement stays a human step. Guide: [docs/claude-skill.md](docs/claude-skill.md).
+
 ## Daily path
 
 ```bash

@@ -264,6 +264,21 @@ Deep dive: [docs/operator-loop.md](./docs/operator-loop.md) · [docs/commands.md
 
 ---
 
+
+## Claude Code (additive host)
+
+Hermes remains the primary skill host. Claude Code is optional:
+
+```bash
+bash install.sh --claude
+export HYPERLEX_SKILL_DIR="${HOME}/.claude/skills/hyperlex"
+export HLX="python3 $HYPERLEX_SKILL_DIR/scripts/hyperlex.py"
+$HLX doctor   # CLAUDE_OK / CLAUDE_MISSING (missing does not fail)
+```
+
+See [`docs/claude-skill.md`](docs/claude-skill.md) and [`docs/claude-runtime-contract.md`](docs/claude-runtime-contract.md).
+
+
 ## CLI surface
 
 **Prefer**