diff --git a/.claude-plugin/plugin.json b/.claude-plugin/plugin.json
new file mode 100644
index 0000000..7d656b6
--- /dev/null
+++ b/.claude-plugin/plugin.json
@@ -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"
+  ]
+}
diff --git a/.claude/skills/hyperlex-analyze/SKILL.md b/.claude/skills/hyperlex-analyze/SKILL.md
new file mode 100644
index 0000000..b1f4048
--- /dev/null
+++ b/.claude/skills/hyperlex-analyze/SKILL.md
@@ -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 "$ARGUMENTS" --route offline
+```
+
+## Expected
+
+JSON result. `provenance.brier` is `null`. Lineage / virality / mutation_trace only when the CLI attaches them.
+
+## Fail-closed
+
+Do not invent labels or Brier. If the file is missing, say so. Do not settle.
diff --git a/.claude/skills/hyperlex-demo/SKILL.md b/.claude/skills/hyperlex-demo/SKILL.md
new file mode 100644
index 0000000..0c44440
--- /dev/null
+++ b/.claude/skills/hyperlex-demo/SKILL.md
@@ -0,0 +1,29 @@
+---
+name: hyperlex-demo
+description: Run the Hyperlex offline demo and summarize receipt labels. No API keys. Never invent Brier.
+when_to_use: User asks for a Hyperlex demo, first success path, or a sample receipt.
+allowed-tools: Bash Read
+---
+
+# /hyperlex-demo
+
+Run the offline demo. Summarize labels from the CLI output. Do not invent scores.
+
+```bash
+: "${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.
diff --git a/.claude/skills/hyperlex-pending/SKILL.md b/.claude/skills/hyperlex-pending/SKILL.md
new file mode 100644
index 0000000..b939415
--- /dev/null
+++ b/.claude/skills/hyperlex-pending/SKILL.md
@@ -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
+$HLX pending --limit 20
+```
+
+## Expected
+
+JSON list of open forecast ids (default log: `~/.hyperlex/score_log.jsonl`). Empty list is valid.
+
+## Fail-closed
+
+Do not invent forecast ids. Do not compute Brier here. Next step is human: `/hyperlex-settle` after they choose TRUE|FALSE|VOID|CONFLICT.
diff --git a/.claude/skills/hyperlex-scan/SKILL.md b/.claude/skills/hyperlex-scan/SKILL.md
new file mode 100644
index 0000000..44d9dc7
--- /dev/null
+++ b/.claude/skills/hyperlex-scan/SKILL.md
@@ -0,0 +1,28 @@
+---
+name: hyperlex-scan
+description: LIVE_EMERGENCE_SCAN helper. Pass a query (or comma-separated queries). Offline-first.
+when_to_use: User wants a Hyperlex scan, live-emergence scan, or multi-query monitor run.
+allowed-tools: Bash Read
+arguments: [query]
+---
+
+# /hyperlex-scan
+
+Multi-query LIVE_EMERGENCE_SCAN. Prefer `--route offline` unless the user allows network.
+
+```bash
+: "${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.
diff --git a/.claude/skills/hyperlex-settle/SKILL.md b/.claude/skills/hyperlex-settle/SKILL.md
new file mode 100644
index 0000000..974f8f7
--- /dev/null
+++ b/.claude/skills/hyperlex-settle/SKILL.md
@@ -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
+$HLX score-series --mean-shift --verify-chain
+```
+
+## Expected
+
+Settle appends the score log. `score-series` is the only place a numeric Brier is real. Empty series → `NOT_COMPUTABLE`.
+
+## Fail-closed
+
+If the operator has not chosen a decision, stop. Coach the flags. Do not auto-settle. Do not guess outcomes.
diff --git a/.claude/skills/hyperlex-wizard/SKILL.md b/.claude/skills/hyperlex-wizard/SKILL.md
new file mode 100644
index 0000000..158b429
--- /dev/null
+++ b/.claude/skills/hyperlex-wizard/SKILL.md
@@ -0,0 +1,26 @@
+---
+name: hyperlex-wizard
+description: Week-one Hyperlex onboarding via wizard --auto. Offline. Never auto-settles.
+when_to_use: User is new to Hyperlex, asks to get started, or wants the guided operator path.
+allowed-tools: Bash Read
+---
+
+# /hyperlex-wizard
+
+Week-one guided path. Prefer `--auto` in Claude (non-interactive).
+
+```bash
+: "${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 `$HLX settle`. Do not invent Brier.
diff --git a/.claude/skills/hyperlex/SKILL.md b/.claude/skills/hyperlex/SKILL.md
new file mode 100644
index 0000000..1ed2af1
--- /dev/null
+++ b/.claude/skills/hyperlex/SKILL.md
@@ -0,0 +1,43 @@
+---
+name: hyperlex
+description: >
+  Main Hyperlex contract in Claude Code. Use for slang, memetics, lineage,
+  virality, hyperstition, receipts, forecasts, and settlement coaching.
+  Prefer the CLI. Not for jailbreak / wrap composition.
+when_to_use: >
+  User asks for Hyperlex, slang analysis, memetic emergence, lineage,
+  Brier settlement, or cultural-signal receipts.
+allowed-tools: Bash Read Grep
+---
+
+# Hyperlex (project skill)
+
+Read the repo-root [SKILL.md](../../../SKILL.md) for the full Hermes + Claude
+contract. This file is a thin pointer so Claude Code can invoke `/hyperlex`
+when this checkout is the project. Do not duplicate engine logic here.
+
+## Invoke
+
+```bash
+export HYPERLEX_SKILL_DIR="${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.
diff --git a/.gitignore b/.gitignore
index adec837..6adaf42 100644
--- a/.gitignore
+++ b/.gitignore
@@ -38,6 +38,7 @@ build/
 !docs/command-router.v1.json
 !data/backfill/**/*.json
 !data/phylogeny/**/*.json
+!.claude-plugin/plugin.json
 
 # Temp / build output
 tmp/
diff --git a/CHANGELOG.md b/CHANGELOG.md
index 55f1646..99e2a7b 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -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.
diff --git a/CLAUDE.md b/CLAUDE.md
index 0390e52..c1c0f5d 100644
--- a/CLAUDE.md
+++ b/CLAUDE.md
@@ -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.
diff --git a/QUICKSTART.md b/QUICKSTART.md
index 68ccfa1..f628d0c 100644
--- a/QUICKSTART.md
+++ b/QUICKSTART.md
@@ -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
diff --git a/README.md b/README.md
index 4d2189e..eebb920 100644
--- a/README.md
+++ b/README.md
@@ -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**
diff --git a/SKILL.md b/SKILL.md
index d0e24b0..7d318ae 100644
--- a/SKILL.md
+++ b/SKILL.md
@@ -10,6 +10,14 @@ description: >
   ai-native slang, brainrot, and receipt-backed cultural signal scans.
   Not for general web research (use agent-reach), product audits (neon-genie),
   cinematic work (kubrick), or jailbreak / wrap composition.
+when_to_use: >
+  Invoke for slang detection, memetic emergence, lineage matching,
+  hyperstition or virality scoring, forecast settlement, or mutation
+  traces. Prefer the Hyperlex CLI over inventing scores.
+allowed-tools:
+  - Bash
+  - Read
+  - Grep
 version: 0.4.2
 author: Applied Alchemy Labs / Hermes
 license: LicenseRef-Zero-State-Proprietary-1.0
@@ -30,6 +38,11 @@ metadata:
       - Mutation
     category: analysis
     related_skills: []
+  claude:
+    hosts: [claude-code]
+    personal_skill: ~/.claude/skills/hyperlex
+    plugin_dir: ~/.claude/plugins/hyperlex
+    plugin_manifest: .claude-plugin/plugin.json
   openclaw:
     requires:
       bins: [python3]
@@ -115,6 +128,47 @@ python3 "$HOME/.hermes/skills/hyperlex/scripts/hyperlex.py" check
 python3 "$HOME/.hermes/skills/hyperlex/scripts/hyperlex.py" smoke
 ```
 
+Claude Code personal skill (Hermes install still runs; this flag is additive):
+
+```bash
+bash install.sh --claude --dry-run
+bash install.sh --claude
+# ~/.claude/skills/hyperlex  + sibling slash helpers
+# optional plugin tree: bash install.sh --claude-plugin
+```
+
+## Claude Code
+
+Claude is an additional host. Hermes remains the primary skill surface.
+Prefer the Bash CLI. Do not invent scores. Do not auto-settle.
+
+```bash
+export HYPERLEX_SKILL_DIR="${HYPERLEX_SKILL_DIR:-${HERMES_SKILL_DIR:-$HOME/.claude/skills/hyperlex}}"
+export HERMES_SKILL_DIR="${HERMES_SKILL_DIR:-$HYPERLEX_SKILL_DIR}"
+export HLX="${HLX:-python3 $HYPERLEX_SKILL_DIR/scripts/hyperlex.py}"
+# wrapper: bash scripts/claude_hlx.sh
+# after pip: hyperlex
+```
+
+When to call:
+
+| Need | Command |
+|------|---------|
+| First success | `$HLX demo` |
+| Week-one path | `$HLX wizard --auto` |
+| Analyze + receipt | `$HLX pipeline "<term>" --route offline` |
+| Multi-query scan | `$HLX scan --query "<term>" --route offline --receipt --forecasts` |
+| Open forecasts | `$HLX pending` |
+| Close a forecast | `$HLX settle --forecast-id <id> --decision TRUE` after the human chooses |
+
+Slash helpers (project `.claude/skills/` or plugin `commands/`): `/hyperlex`,
+`/hyperlex-demo`, `/hyperlex-wizard`, `/hyperlex-scan`, `/hyperlex-analyze`,
+`/hyperlex-pending`, `/hyperlex-settle`.
+
+Fail-closed: never invent Brier; never auto-settle; repeat OBSERVED / INFERRED /
+SPECULATIVE. Docs: `docs/claude-skill.md`, `CLAUDE.md` (project only — plugin
+context does not load `CLAUDE.md`).
+
 ## Commands (prefer simplified path)
 
 ```bash
@@ -270,10 +324,12 @@ PYTHONPATH="$HERMES_SKILL_DIR/src" python3 -m hyperlex mutation trace "it's givi
 Successful packaging:
 
 - `~/.hermes/skills/hyperlex/SKILL.md` exists
+- optional Claude: `~/.claude/skills/hyperlex/SKILL.md` after `install.sh --claude`
 - `check` returns `"ok": true`
 - `smoke` writes a receipt under `out/smoke/`
 - Open analysis has `"brier": null`
 - Mutation trace packets have `"forecast_eligible": false`
+- `doctor` reports `CLAUDE_OK` or `CLAUDE_MISSING` (missing does not fail the skill)
 
 ## Design references
 
@@ -287,6 +343,7 @@ Successful packaging:
 - `examples/slang-families/` — Mermaid + HTML family diagrams
 - `data/backfill/2026/` — YTD slang packs
 - `references/hermes-runtime-contract.md` — path / authority policy
+- `references/claude-runtime-contract.md` — Claude Code paths / env / CLI
 - `references/source-and-upgrades.md` — install identity, two-rename limits, lock recovery
 
 ## Security
diff --git a/STATUS.md b/STATUS.md
index 63ea392..1ce0377 100644
--- a/STATUS.md
+++ b/STATUS.md
@@ -3,6 +3,7 @@
 **Version:** 0.4.0  
 **Posture:** Hermes skill (Python package repo)  
 **Install:** `bash install.sh` → `~/.hermes/skills/hyperlex`  
+**Claude (optional):** `bash install.sh --claude` → `~/.claude/skills/hyperlex`  
 **Track:** Phases 0–4 complete · Phase 5.0–5.3 · Pages static run history · Hallmark desk UI
 
 ## Health
diff --git a/commands/hyperlex-analyze.md b/commands/hyperlex-analyze.md
new file mode 100644
index 0000000..867f0d3
--- /dev/null
+++ b/commands/hyperlex-analyze.md
@@ -0,0 +1,23 @@
+---
+description: Analyze pasted text or a JSON ingest file with the Hyperlex CLI. Prefer offline route.
+allowed-tools: Bash Read
+---
+
+# /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:-$HOME/.claude/skills/hyperlex}}/scripts/hyperlex.py}"
+$HLX analyze "$ARGUMENTS" --route offline
+$HLX pipeline "$ARGUMENTS" --route offline
+$HLX analyze --input "$ARGUMENTS" --route offline
+```
+
+## Expected
+
+JSON result. `provenance.brier` is `null`. Lineage / virality / mutation_trace only when the CLI attaches them.
+
+## Fail-closed
+
+Do not invent labels or Brier. If the file is missing, say so. Do not settle.
diff --git a/commands/hyperlex-demo.md b/commands/hyperlex-demo.md
new file mode 100644
index 0000000..3e87c61
--- /dev/null
+++ b/commands/hyperlex-demo.md
@@ -0,0 +1,27 @@
+---
+description: Run the Hyperlex offline demo and summarize receipt labels. No API keys. Never invent Brier.
+allowed-tools: Bash Read
+---
+
+# /hyperlex-demo
+
+Run the offline demo. Summarize labels from the CLI output. Do not invent scores.
+
+```bash
+: "${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.
diff --git a/commands/hyperlex-pending.md b/commands/hyperlex-pending.md
new file mode 100644
index 0000000..af7af01
--- /dev/null
+++ b/commands/hyperlex-pending.md
@@ -0,0 +1,22 @@
+---
+description: List open Hyperlex forecasts from the score log. Start of the calibration loop.
+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
+$HLX pending --limit 20
+```
+
+## Expected
+
+JSON list of open forecast ids (default log: `~/.hyperlex/score_log.jsonl`). Empty list is valid.
+
+## Fail-closed
+
+Do not invent forecast ids. Do not compute Brier here. Next step is human: `/hyperlex-settle` after they choose TRUE|FALSE|VOID|CONFLICT.
diff --git a/commands/hyperlex-scan.md b/commands/hyperlex-scan.md
new file mode 100644
index 0000000..548c9ea
--- /dev/null
+++ b/commands/hyperlex-scan.md
@@ -0,0 +1,23 @@
+---
+description: LIVE_EMERGENCE_SCAN helper. Pass a query (or comma-separated queries). Offline-first.
+allowed-tools: Bash Read
+---
+
+# /hyperlex-scan
+
+Multi-query LIVE_EMERGENCE_SCAN. Prefer `--route offline` unless the user allows network.
+
+```bash
+: "${HLX:=python3 ${HYPERLEX_SKILL_DIR:-${HERMES_SKILL_DIR:-$HOME/.claude/skills/hyperlex}}/scripts/hyperlex.py}"
+$HLX scan --query "$ARGUMENTS" --route offline --receipt --forecasts --append-log
+$HLX scan --queries "rizz,locked in" --route offline --receipt --forecasts --append-log
+$HLX scan --config "${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.
diff --git a/commands/hyperlex-settle.md b/commands/hyperlex-settle.md
new file mode 100644
index 0000000..938c36b
--- /dev/null
+++ b/commands/hyperlex-settle.md
@@ -0,0 +1,27 @@
+---
+description: Coach Hyperlex settlement. Human must supply forecast-id and decision. Never invent outcomes or Brier.
+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
+$HLX score-series --mean-shift --verify-chain
+```
+
+## Expected
+
+Settle appends the score log. `score-series` is the only place a numeric Brier is real. Empty series → `NOT_COMPUTABLE`.
+
+## Fail-closed
+
+If the operator has not chosen a decision, stop. Coach the flags. Do not auto-settle. Do not guess outcomes.
diff --git a/commands/hyperlex-wizard.md b/commands/hyperlex-wizard.md
new file mode 100644
index 0000000..4d2a031
--- /dev/null
+++ b/commands/hyperlex-wizard.md
@@ -0,0 +1,24 @@
+---
+description: Week-one Hyperlex onboarding via wizard --auto. Offline. Never auto-settles.
+allowed-tools: Bash Read
+---
+
+# /hyperlex-wizard
+
+Week-one guided path. Prefer `--auto` in Claude (non-interactive).
+
+```bash
+: "${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.
diff --git a/docs/README.md b/docs/README.md
index d5716d1..ceaf187 100644
--- a/docs/README.md
+++ b/docs/README.md
@@ -8,6 +8,7 @@ Hyperlex is a **Hermes skill** (Python package repo). Relevant Abraxas wire shap
 
 ## Core
 - [Hermes skill model](./hermes-skill.md)
+- [Claude Code skill + plugin](./claude-skill.md)
 - [Public API v1 freeze](./api-v1.md)
 - [Brier & Calibration](./brier-calibration.md)
 - [Rune / Signal Relay](./rune-signal-relay.md)
@@ -35,4 +36,5 @@ Hyperlex is a **Hermes skill** (Python package repo). Relevant Abraxas wire shap
 ## References
 - [arXiv papers](../references/arxiv_papers.md)
 - [Hermes runtime contract](../references/hermes-runtime-contract.md)
+- [Claude Code runtime contract](claude-runtime-contract.md)
 
diff --git a/docs/claude-runtime-contract.md b/docs/claude-runtime-contract.md
new file mode 100644
index 0000000..c3acb91
--- /dev/null
+++ b/docs/claude-runtime-contract.md
@@ -0,0 +1,79 @@
+# Hyperlex — Claude Code runtime contract
+
+Published operator contract (MkDocs). The skill-tree copy lives at
+`references/claude-runtime-contract.md`.
+
+## Identity
+
+| Field | Value |
+|-------|--------|
+| Skill / plugin name | `hyperlex` |
+| Personal skill | `~/.claude/skills/hyperlex` |
+| Local plugin dir | `~/.claude/plugins/hyperlex` |
+| Plugin manifest | `.claude-plugin/plugin.json` |
+| Entry contract | root `SKILL.md` (single-skill plugin layout) |
+| CLI | `scripts/hyperlex.py` |
+| Wrapper | `scripts/claude_hlx.sh` |
+| Package | `src/hyperlex/` |
+
+Claude Code loads a root `SKILL.md` as one skill when the plugin has no
+`skills/` directory and no `skills` manifest field. Slash helpers are project
+skills under `.claude/skills/` and plugin commands under `commands/`. Do not
+copy the engine into `skills/hyperlex/`.
+
+`CLAUDE.md` is project context when this repo is the working directory. It is
+not plugin context.
+
+## Paths
+
+```text
+Personal skill (~/.claude/skills/hyperlex/):
+  SKILL.md
+  scripts/hyperlex.py
+  scripts/claude_hlx.sh
+  src/hyperlex/
+
+Local plugin (~/.claude/plugins/hyperlex/):
+  .claude-plugin/plugin.json
+  SKILL.md
+  commands/hyperlex-*.md
+  scripts/  src/
+
+Project checkout:
+  CLAUDE.md
+  .claude/skills/<helper>/SKILL.md
+```
+
+Operator data is unchanged (outside the skill tree):
+
+```text
+~/.hyperlex/receipts/
+~/.hyperlex/score_log.jsonl
+```
+
+## Environment
+
+| Variable | Role |
+|----------|------|
+| `HLX` | CLI invocation string (`python3 …/scripts/hyperlex.py`) |
+| `HYPERLEX_SKILL_DIR` | Claude-preferred skill root |
+| `CLAUDE_SKILL_DIR` | Optional alias for `HYPERLEX_SKILL_DIR` |
+| `HERMES_SKILL_DIR` | Hermes name for the same tree; Claude may reuse it |
+
+`scripts/claude_hlx.sh` resolves the skill dir in this order:
+`HYPERLEX_SKILL_DIR`, `CLAUDE_SKILL_DIR`, `HERMES_SKILL_DIR`, this checkout,
+`~/.claude/skills/hyperlex`, `~/.claude/plugins/hyperlex`,
+`~/.hermes/skills/hyperlex`. Then it execs `scripts/hyperlex.py`.
+
+## Execution rules
+
+1. Prefer Bash CLI: `$HLX …` or `bash scripts/claude_hlx.sh …`.
+2. Baseline work uses `--route offline` (no network, no API keys).
+3. Never invent numeric Brier. Open analysis keeps `provenance.brier` null.
+4. Never auto-settle. Ask for TRUE|FALSE|VOID|CONFLICT, then run `settle`.
+5. `doctor` reports `CLAUDE_OK` or `CLAUDE_MISSING`. Missing does not fail Hermes.
+
+## Authority
+
+Same classes as the Hermes runtime contract (`references/hermes-runtime-contract.md`):
+OBSERVED, INFERRED, SPECULATIVE, NOT_COMPUTABLE, operator settlement.
diff --git a/docs/claude-skill.md b/docs/claude-skill.md
new file mode 100644
index 0000000..c13c8b3
--- /dev/null
+++ b/docs/claude-skill.md
@@ -0,0 +1,93 @@
+# Hyperlex in Claude Code
+
+Hermes is the primary skill host. Claude Code is an additional host. Same CLI,
+same settlement rule, same receipts. This is dual-runtime packaging — not a
+rewrite.
+
+## Two install shapes
+
+| Shape | Path | What Claude loads |
+|-------|------|-------------------|
+| **Personal skill** | `~/.claude/skills/hyperlex/` | Directory `SKILL.md` + bundled `scripts/` / `src/` for offline CLI |
+| **Local plugin** | `~/.claude/plugins/hyperlex/` | `.claude-plugin/plugin.json`, root `SKILL.md`, slash `commands/` |
+
+The repo itself is already a single-skill plugin: root `SKILL.md` plus
+`.claude-plugin/plugin.json`. You can point Claude Code at this checkout
+instead of copying it.
+
+Thin `hyperlex init --target claude` writes only `SKILL.md` and expects
+`hyperlex` on PATH. Prefer `install.sh --claude` when you want the offline CLI
+tree without a pip install.
+
+## Personal skill (recommended for operators)
+
+```bash
+bash install.sh --claude --dry-run
+bash install.sh --claude
+export HYPERLEX_SKILL_DIR="${HOME}/.claude/skills/hyperlex"
+export HERMES_SKILL_DIR="${HERMES_SKILL_DIR:-$HYPERLEX_SKILL_DIR}"
+export HLX="python3 $HYPERLEX_SKILL_DIR/scripts/hyperlex.py"
+$HLX check && $HLX doctor && $HLX demo
+```
+
+`--claude` is **additive**. Hermes still installs to `~/.hermes/skills/hyperlex`
+unless you pass `--target`. Sibling slash helpers land next to the personal
+skill:
+
+```text
+~/.claude/skills/hyperlex/          # main contract + CLI
+~/.claude/skills/hyperlex-demo/
+~/.claude/skills/hyperlex-wizard/
+~/.claude/skills/hyperlex-scan/
+~/.claude/skills/hyperlex-analyze/
+~/.claude/skills/hyperlex-pending/
+~/.claude/skills/hyperlex-settle/
+```
+
+Reload Claude Code so it sees the new skills.
+
+## Local plugin dir
+
+```bash
+bash install.sh --claude-plugin --dry-run
+bash install.sh --claude-plugin
+# tree: ~/.claude/plugins/hyperlex
+# enable via Claude plugin UI, or: claude plugin add ~/.claude/plugins/hyperlex
+```
+
+Plugin slash commands live in repo `commands/` (`/hyperlex-demo`, …). The main
+`/hyperlex` skill is the root `SKILL.md` (single-skill layout — no duplicate
+engine under `skills/hyperlex/`).
+
+You can also add this git checkout as a plugin path. `CLAUDE.md` at the repo
+root is **project** context only; Claude does not load it as plugin context.
+
+## First success (no Anthropic API)
+
+```bash
+$HLX demo
+$HLX wizard --auto
+```
+
+Expect `ok: true` and `brier: null`. No paid keys. Settlement stays a human
+step: `pending` → operator decision → `settle` → `score-series`.
+
+## Differences from Hermes
+
+| | Hermes | Claude Code |
+|--|--------|-------------|
+| Default skill path | `~/.hermes/skills/hyperlex` | `~/.claude/skills/hyperlex` |
+| Env name | `HERMES_SKILL_DIR` | `HYPERLEX_SKILL_DIR` (may reuse Hermes) |
+| Wrapper | `python3 $HERMES_SKILL_DIR/scripts/hyperlex.py` | `scripts/claude_hlx.sh` |
+| Discovery | Hermes scans `SKILL.md` | Personal skills + optional plugin |
+| Slash helpers | Hermes procedure in `SKILL.md` | `.claude/skills/` and `commands/` |
+| Contract | Same `SKILL.md` | Same `SKILL.md` + Claude section |
+
+Fail-closed rules do not change: no invented Brier, no auto-settle, no
+phenomenology claims.
+
+## See also
+
+- [Claude runtime contract](claude-runtime-contract.md)
+- [Hermes skill model](hermes-skill.md)
+- [Operator loop](operator-loop.md)
diff --git a/docs/commands.md b/docs/commands.md
index 99aabce..7d684a5 100644
--- a/docs/commands.md
+++ b/docs/commands.md
@@ -4,7 +4,7 @@ Hyperlex has many subcommands. Prefer this map. Full list: `$HLX commands` (JSON
 or `python3 scripts/hyperlex.py -h`.
 
 ```bash
-HLX="python3 ${HERMES_SKILL_DIR:-$HOME/.hermes/skills/hyperlex}/scripts/hyperlex.py"
+HLX="python3 ${HYPERLEX_SKILL_DIR:-${HERMES_SKILL_DIR:-$HOME/.hermes/skills/hyperlex}}/scripts/hyperlex.py"
 ```
 
 ## Daily ops (automatic backend)
diff --git a/docs/hermes-skill.md b/docs/hermes-skill.md
index 61e6e55..4466bf7 100644
--- a/docs/hermes-skill.md
+++ b/docs/hermes-skill.md
@@ -35,6 +35,9 @@ $HLX check && $HLX doctor && $HLX commands
 $HLX wizard --auto
 ```
 
+Claude Code is an additional host (same CLI, same settlement rule):
+[claude-skill.md](claude-skill.md). `--claude` on `install.sh` is additive.
+
 The **wizard** is the preferred first-run Hermes procedure: offline guided path
 (env → doctor → demo → first pipeline → calibration coach). It never auto-settles;
 settlement stays a human operator step (`pending` → `settle` → `score-series`).
diff --git a/docs/install-package.md b/docs/install-package.md
index a54fbe0..4655393 100644
--- a/docs/install-package.md
+++ b/docs/install-package.md
@@ -23,3 +23,12 @@ It does **not** rsync `src/` or `scripts/hyperlex.py`. The agent runs `hyperlex
 Reload the agent session after init.
 
 Legacy full-tree install remains `bash install.sh` for hosts that still want the fat skill CLI.
+Claude Code full tree (SKILL.md + scripts/src, not the thin init wire):
+
+```bash
+bash install.sh --claude --dry-run
+bash install.sh --claude            # ~/.claude/skills/hyperlex
+bash install.sh --claude-plugin     # ~/.claude/plugins/hyperlex
+```
+
+See [claude-skill.md](claude-skill.md).
diff --git a/docs/start/quickstart.md b/docs/start/quickstart.md
index 20878b2..c448440 100644
--- a/docs/start/quickstart.md
+++ b/docs/start/quickstart.md
@@ -45,6 +45,18 @@ $HLX demo
 $HLX wizard --auto
 ```
 
+## 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
+```
+
+Same offline first success. Hermes remains the default `install.sh` target;
+`--claude` is additive. Details: [Claude skill](../claude-skill.md).
+
 ## Committed sample output
 
 See [`examples/quickstart/`](https://github.com/scrimshawlife-ctrl/Hyperlex/tree/main/examples/quickstart)
diff --git a/hyperlex.manifest.yaml b/hyperlex.manifest.yaml
index 75a300c..e6f9027 100644
--- a/hyperlex.manifest.yaml
+++ b/hyperlex.manifest.yaml
@@ -1,5 +1,5 @@
 name: hyperlex
-version: 0.4.0
+version: 0.4.2
 display_name: Hyperlex
 kind: hermes-openclaw-skill
 description: >
@@ -7,8 +7,8 @@ description: >
   tracking, virality analysis, lineage matching, settled Brier calibration,
   and slang mutation-operator detection on attested text.
 license: LicenseRef-Zero-State-Proprietary-1.0
-homepage: https://github.com/scrimshawlife-ctrl/Hyperlex
-repository: https://github.com/scrimshawlife-ctrl/Hyperlex
+homepage: https://github.com/Zero-State-LLC/Hyperlex
+repository: https://github.com/Zero-State-LLC/Hyperlex
 
 skill_entry: SKILL.md
 cli_entry: scripts/hyperlex.py
@@ -19,10 +19,13 @@ package_root: src/hyperlex
 hosts:
   - hermes
   - openclaw
+  - claude-code
 
 install:
   hermes_path: ~/.hermes/skills/hyperlex
   openclaw_path: ~/.openclaw/skills/hyperlex
+  claude_skill_path: ~/.claude/skills/hyperlex
+  claude_plugin_path: ~/.claude/plugins/hyperlex
   atomic: true
   supports_dry_run: true
   supports_rollback: true
diff --git a/install.sh b/install.sh
index 0b0b407..af4175e 100755
--- a/install.sh
+++ b/install.sh
@@ -1,11 +1,13 @@
 #!/usr/bin/env bash
-# Hyperlex — Hermes skill installer
+# Hyperlex — Hermes skill installer (Claude Code is an additional host)
 # Usage:
 #   ./install.sh
 #   ./install.sh --dry-run
 #   ./install.sh --target DIR
 #   ./install.sh --rollback
 #   ./install.sh --openclaw
+#   ./install.sh --claude
+#   ./install.sh --claude-plugin
 #   ./install.sh --version
 set -euo pipefail
 
@@ -13,6 +15,16 @@ ROOT="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
 HERMES_ROOT="${HERMES_HOME:-$HOME/.hermes}"
 DEFAULT_TARGET="${HERMES_ROOT}/skills/hyperlex"
 OPENCLAW_TARGET="${HOME}/.openclaw/skills/hyperlex"
+CLAUDE_SKILL_TARGET="${HOME}/.claude/skills/hyperlex"
+CLAUDE_PLUGIN_TARGET="${HOME}/.claude/plugins/hyperlex"
+CLAUDE_HELPERS=(
+  hyperlex-demo
+  hyperlex-wizard
+  hyperlex-scan
+  hyperlex-analyze
+  hyperlex-pending
+  hyperlex-settle
+)
 
 VERSION="$(tr -d '[:space:]' < "${ROOT}/VERSION" 2>/dev/null || echo "0.0.0")"
 
@@ -20,6 +32,8 @@ DRY_RUN=0
 ROLLBACK=0
 ALLOW_OUTSIDE_HOME=0
 INSTALL_OPENCLAW=0
+INSTALL_CLAUDE=0
+INSTALL_CLAUDE_PLUGIN=0
 SKIP_SMOKE=0
 TARGET="$DEFAULT_TARGET"
 
@@ -29,15 +43,25 @@ Hyperlex Hermes skill installer v${VERSION}
 
 Usage: ./install.sh [options]
 
+Hermes remains the default target. Claude flags are additive extra hosts.
+
 Options:
   --dry-run               Show actions without writing
   --target DIR            Install to DIR (default: ${DEFAULT_TARGET})
   --rollback              Restore most recent target-keyed backup
   --openclaw              Also install to ~/.openclaw/skills/hyperlex
+  --claude                Also install personal skill to ~/.claude/skills/hyperlex
+                          plus slash helpers (hyperlex-demo, …) as sibling skills
+  --claude-plugin         Also install plugin tree to ~/.claude/plugins/hyperlex
   --skip-smoke            Skip staged check/smoke (marks UNVERIFIED)
   --allow-outside-home    Permit --target outside \$HOME
   --version               Print version and exit
   -h, --help              Show this help
+
+Claude paths:
+  Personal skill:  ~/.claude/skills/hyperlex/   (SKILL.md + scripts/src)
+  Local plugin:    ~/.claude/plugins/hyperlex/  (.claude-plugin/ + commands/)
+  Project helpers: .claude/skills/ in this repo (when the checkout is the project)
 EOF
 }
 
@@ -62,6 +86,8 @@ while [[ $# -gt 0 ]]; do
     --dry-run) DRY_RUN=1; shift ;;
     --rollback) ROLLBACK=1; shift ;;
     --openclaw) INSTALL_OPENCLAW=1; shift ;;
+    --claude) INSTALL_CLAUDE=1; shift ;;
+    --claude-plugin) INSTALL_CLAUDE_PLUGIN=1; shift ;;
     --skip-smoke) SKIP_SMOKE=1; shift ;;
     --allow-outside-home) ALLOW_OUTSIDE_HOME=1; shift ;;
     --version) echo "$VERSION"; exit 0 ;;
@@ -116,7 +142,9 @@ validate_source() {
     "VERSION"
     "scripts/hyperlex.py"
     "scripts/hlx-mutation"
+    "scripts/claude_hlx.sh"
     "scripts/install_transaction.py"
+    ".claude-plugin/plugin.json"
     "src/hyperlex/__init__.py"
     "src/hyperlex/calibration/scoring.py"
     "src/hyperlex/schemas/result.v1.schema.json"
@@ -149,6 +177,40 @@ run_transaction() {
   python3 "${ROOT}/scripts/install_transaction.py" "$ROOT" "$dest" hyperlex "${check_args[@]}"
 }
 
+install_extra_host() {
+  local dest="$1"
+  local label="$2"
+  local _saved="$TARGET"
+  TARGET="$dest"
+  validate_target
+  if [[ $DRY_RUN -eq 1 ]]; then
+    log "DRY RUN: would also install ${label} to ${TARGET}"
+    TARGET="$_saved"
+    return 0
+  fi
+  run_transaction "$TARGET"
+  TARGET="$_saved"
+}
+
+copy_claude_helpers() {
+  local dest_root="${HOME}/.claude/skills"
+  local name src dest
+  for name in "${CLAUDE_HELPERS[@]}"; do
+    src="${ROOT}/.claude/skills/${name}/SKILL.md"
+    dest="${dest_root}/${name}/SKILL.md"
+    if [[ ! -f "$src" ]]; then
+      warn "Claude helper missing in source: ${src}"
+      continue
+    fi
+    if [[ $DRY_RUN -eq 1 ]]; then
+      log "DRY RUN: would install Claude helper ${name} → ${dest}"
+      continue
+    fi
+    mkdir -p "$(dirname "$dest")"
+    cp -f "$src" "$dest"
+  done
+}
+
 do_rollback() {
   validate_target
   if [[ $DRY_RUN -eq 1 ]]; then
@@ -174,17 +236,25 @@ if [[ $DRY_RUN -eq 1 ]]; then
   log "DRY RUN: two-rename activation is not crash-atomic; locks are never auto-reclaimed"
   [[ -d "$TARGET" ]] && log "DRY RUN: would publish a target-keyed backup under ${HERMES_ROOT}/backups/hyperlex/"
   [[ $INSTALL_OPENCLAW -eq 1 ]] && log "DRY RUN: would also install to ${OPENCLAW_TARGET}"
+  [[ $INSTALL_CLAUDE -eq 1 ]] && log "DRY RUN: would also install Claude personal skill to ${CLAUDE_SKILL_TARGET}"
+  [[ $INSTALL_CLAUDE -eq 1 ]] && copy_claude_helpers
+  [[ $INSTALL_CLAUDE_PLUGIN -eq 1 ]] && log "DRY RUN: would also install Claude plugin dir to ${CLAUDE_PLUGIN_TARGET}"
   exit 0
 fi
 
 run_transaction "$TARGET"
 
 if [[ $INSTALL_OPENCLAW -eq 1 ]]; then
-  _saved="$TARGET"
-  TARGET="$OPENCLAW_TARGET"
-  validate_target
-  run_transaction "$TARGET"
-  TARGET="$_saved"
+  install_extra_host "$OPENCLAW_TARGET" "OpenClaw"
+fi
+
+if [[ $INSTALL_CLAUDE -eq 1 ]]; then
+  install_extra_host "$CLAUDE_SKILL_TARGET" "Claude personal skill"
+  copy_claude_helpers
+fi
+
+if [[ $INSTALL_CLAUDE_PLUGIN -eq 1 ]]; then
+  install_extra_host "$CLAUDE_PLUGIN_TARGET" "Claude plugin"
 fi
 
 echo ""
@@ -195,11 +265,22 @@ else
 fi
 echo "  Hermes:  ${TARGET}"
 [[ $INSTALL_OPENCLAW -eq 1 ]] && echo "  OpenClaw: ${OPENCLAW_TARGET}"
+[[ $INSTALL_CLAUDE -eq 1 ]] && echo "  Claude:   ${CLAUDE_SKILL_TARGET}"
+[[ $INSTALL_CLAUDE_PLUGIN -eq 1 ]] && echo "  Plugin:   ${CLAUDE_PLUGIN_TARGET}"
 echo ""
 echo "Next:"
 echo "  export HERMES_SKILL_DIR=\"${TARGET}\""
 echo "  python3 \"\$HERMES_SKILL_DIR/scripts/hyperlex.py\" check"
 echo "  python3 \"\$HERMES_SKILL_DIR/scripts/hyperlex.py\" pipeline \"rizz\" --route offline"
 echo "  python3 \"\$HERMES_SKILL_DIR/scripts/hlx-mutation\" trace \"it's giving mid rizz\""
+if [[ $INSTALL_CLAUDE -eq 1 ]]; then
+  echo "  export HYPERLEX_SKILL_DIR=\"${CLAUDE_SKILL_TARGET}\""
+  echo "  export HLX=\"python3 \$HYPERLEX_SKILL_DIR/scripts/hyperlex.py\""
+  echo "  bash \"\$HYPERLEX_SKILL_DIR/scripts/claude_hlx.sh\" check"
+  echo "  # Reload Claude Code so it sees ~/.claude/skills/hyperlex"
+fi
+if [[ $INSTALL_CLAUDE_PLUGIN -eq 1 ]]; then
+  echo "  # Enable the local plugin, or: claude plugin add ${CLAUDE_PLUGIN_TARGET}"
+fi
 echo "  # Reload Hermes skills if the agent is already running"
 echo ""
diff --git a/mkdocs.yml b/mkdocs.yml
index c3dad2a..7269879 100644
--- a/mkdocs.yml
+++ b/mkdocs.yml
@@ -71,6 +71,8 @@ nav:
       - Cron / live emergence: cron-live-emergence.md
       - Case studies: case-studies.md
       - Hermes skill: hermes-skill.md
+      - Claude Code skill: claude-skill.md
+      - Claude runtime contract: claude-runtime-contract.md
   - Core Reference:
       - Architecture: architecture.md
       - Spec: spec.md
diff --git a/references/claude-runtime-contract.md b/references/claude-runtime-contract.md
new file mode 100644
index 0000000..64a917f
--- /dev/null
+++ b/references/claude-runtime-contract.md
@@ -0,0 +1,78 @@
+# Hyperlex — Claude Code runtime contract
+
+Skill-tree copy. Published docs page: `docs/claude-runtime-contract.md`.
+
+## Identity
+
+| Field | Value |
+|-------|--------|
+| Skill / plugin name | `hyperlex` |
+| Personal skill | `~/.claude/skills/hyperlex` |
+| Local plugin dir | `~/.claude/plugins/hyperlex` |
+| Plugin manifest | `.claude-plugin/plugin.json` |
+| Entry contract | root `SKILL.md` (single-skill plugin layout) |
+| CLI | `scripts/hyperlex.py` |
+| Wrapper | `scripts/claude_hlx.sh` |
+| Package | `src/hyperlex/` |
+
+Claude Code loads a root `SKILL.md` as one skill when the plugin has no
+`skills/` directory and no `skills` manifest field. Slash helpers are project
+skills under `.claude/skills/` and plugin commands under `commands/`. Do not
+copy the engine into `skills/hyperlex/`.
+
+`CLAUDE.md` is project context when this repo is the working directory. It is
+not plugin context.
+
+## Paths
+
+```text
+Personal skill (~/.claude/skills/hyperlex/):
+  SKILL.md
+  scripts/hyperlex.py
+  scripts/claude_hlx.sh
+  src/hyperlex/
+
+Local plugin (~/.claude/plugins/hyperlex/):
+  .claude-plugin/plugin.json
+  SKILL.md
+  commands/hyperlex-*.md
+  scripts/  src/
+
+Project checkout:
+  CLAUDE.md
+  .claude/skills/<helper>/SKILL.md
+```
+
+Operator data is unchanged (outside the skill tree):
+
+```text
+~/.hyperlex/receipts/
+~/.hyperlex/score_log.jsonl
+```
+
+## Environment
+
+| Variable | Role |
+|----------|------|
+| `HLX` | CLI invocation string (`python3 …/scripts/hyperlex.py`) |
+| `HYPERLEX_SKILL_DIR` | Claude-preferred skill root |
+| `CLAUDE_SKILL_DIR` | Optional alias for `HYPERLEX_SKILL_DIR` |
+| `HERMES_SKILL_DIR` | Hermes name for the same tree; Claude may reuse it |
+
+`scripts/claude_hlx.sh` resolves the skill dir in this order:
+`HYPERLEX_SKILL_DIR`, `CLAUDE_SKILL_DIR`, `HERMES_SKILL_DIR`, this checkout,
+`~/.claude/skills/hyperlex`, `~/.claude/plugins/hyperlex`,
+`~/.hermes/skills/hyperlex`. Then it execs `scripts/hyperlex.py`.
+
+## Execution rules
+
+1. Prefer Bash CLI: `$HLX …` or `bash scripts/claude_hlx.sh …`.
+2. Baseline work uses `--route offline` (no network, no API keys).
+3. Never invent numeric Brier. Open analysis keeps `provenance.brier` null.
+4. Never auto-settle. Ask for TRUE\|FALSE\|VOID\|CONFLICT, then run `settle`.
+5. `doctor` reports `CLAUDE_OK` or `CLAUDE_MISSING`. Missing does not fail Hermes.
+
+## Authority
+
+Same classes as [hermes-runtime-contract.md](hermes-runtime-contract.md):
+OBSERVED, INFERRED, SPECULATIVE, NOT_COMPUTABLE, operator settlement.
diff --git a/scripts/claude_hlx.sh b/scripts/claude_hlx.sh
new file mode 100755
index 0000000..40ad6b4
--- /dev/null
+++ b/scripts/claude_hlx.sh
@@ -0,0 +1,38 @@
+#!/usr/bin/env bash
+# Claude Code operator entry — set skill dir + HLX, then exec hyperlex.py.
+# Mirrors the Hermes operator path. Offline. No network. No secrets.
+set -euo pipefail
+
+_here="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+_root="$(cd "${_here}/.." && pwd)"
+
+_resolve_skill_dir() {
+  local candidate
+  for candidate in \
+    "${HYPERLEX_SKILL_DIR:-}" \
+    "${CLAUDE_SKILL_DIR:-}" \
+    "${HERMES_SKILL_DIR:-}" \
+    "${_root}" \
+    "${HOME}/.claude/skills/hyperlex" \
+    "${HOME}/.claude/plugins/hyperlex" \
+    "${HOME}/.hermes/skills/hyperlex"
+  do
+    [[ -n "${candidate}" ]] || continue
+    if [[ -f "${candidate}/scripts/hyperlex.py" ]]; then
+      printf '%s\n' "${candidate}"
+      return 0
+    fi
+  done
+  return 1
+}
+
+SKILL_DIR="$(_resolve_skill_dir)" || {
+  printf 'error: Hyperlex CLI not found. Run: bash install.sh --claude\n' >&2
+  exit 1
+}
+
+export HYPERLEX_SKILL_DIR="${SKILL_DIR}"
+export HERMES_SKILL_DIR="${HERMES_SKILL_DIR:-${SKILL_DIR}}"
+export HLX="${HLX:-python3 ${SKILL_DIR}/scripts/hyperlex.py}"
+
+exec python3 "${SKILL_DIR}/scripts/hyperlex.py" "$@"
diff --git a/scripts/hyperlex.py b/scripts/hyperlex.py
index a20689c..0a26ad7 100755
--- a/scripts/hyperlex.py
+++ b/scripts/hyperlex.py
@@ -100,6 +100,34 @@ def _check(condition: bool, name: str, message_ok: str, message_fail: str) -> _C
     return _Check(name=name, ok=bool(condition), message=message_ok if condition else message_fail)
 
 
+def _claude_host_check() -> _Check:
+    """Report Claude Code personal-skill / plugin install. Absence is not a failure."""
+    import os
+
+    home = Path.home()
+    personal = home / ".claude" / "skills" / "hyperlex"
+    plugin = home / ".claude" / "plugins" / "hyperlex"
+    found: List[str] = []
+    if (personal / "SKILL.md").is_file():
+        kind = "personal" if (personal / "scripts" / "hyperlex.py").is_file() else "personal_thin"
+        found.append(f"{kind}={personal}")
+    if (plugin / ".claude-plugin" / "plugin.json").is_file() or (plugin / "SKILL.md").is_file():
+        found.append(f"plugin={plugin}")
+    env_raw = os.environ.get("HYPERLEX_SKILL_DIR") or os.environ.get("CLAUDE_SKILL_DIR") or ""
+    if env_raw:
+        env_path = Path(env_raw).expanduser()
+        if (env_path / "scripts" / "hyperlex.py").is_file() or (env_path / "SKILL.md").is_file():
+            found.append(f"env={env_path}")
+    if found:
+        return _check(True, "claude_host", "CLAUDE_OK: " + "; ".join(found), "")
+    return _check(
+        True,
+        "claude_host",
+        "CLAUDE_MISSING: ~/.claude/skills/hyperlex not installed (optional host)",
+        "",
+    )
+
+
 def _import_hyperlex():
     try:
         import importlib
@@ -224,6 +252,8 @@ def cmd_doctor(_args: argparse.Namespace) -> int:
         "mkdocs.yml",
         "STATUS.md",
         "src/hyperlex/receipt/stats.py",
+        ".claude-plugin/plugin.json",
+        "scripts/claude_hlx.sh",
     ]
     for rel in required_files:
         p = ROOT / rel
@@ -351,6 +381,7 @@ def cmd_doctor(_args: argparse.Namespace) -> int:
     checks.append(
         _check(True, "operator_home", f"~/.hyperlex exists={home_hx.is_dir()} path={home_hx}", "")
     )
+    checks.append(_claude_host_check())
 
     ok = all(c.ok for c in checks)
     _emit({
@@ -2322,7 +2353,9 @@ def _build_parser() -> argparse.ArgumentParser:
     check_parser = subparsers.add_parser("check", help="Validate package and manifest readiness")
     check_parser.set_defaults(func=cmd_check)
 
-    doctor_parser = subparsers.add_parser("doctor", help="Deep Hermes-skill health check")
+    doctor_parser = subparsers.add_parser(
+        "doctor", help="Deep skill health check (Hermes + optional Claude host)"
+    )
     doctor_parser.set_defaults(func=cmd_doctor)
 
     sources_parser = subparsers.add_parser(
diff --git a/tests/test_claude_host.py b/tests/test_claude_host.py
new file mode 100644
index 0000000..2f49039
--- /dev/null
+++ b/tests/test_claude_host.py
@@ -0,0 +1,179 @@
+"""Claude Code packaging: install dry-run + doctor host detection.
+
+Offline only. Does not call Anthropic. Hermes dry-run must stay intact.
+"""
+
+from __future__ import annotations
+
+import json
+import os
+import subprocess
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+SCRIPT = ROOT / "scripts" / "hyperlex.py"
+INSTALL = ROOT / "install.sh"
+
+
+def _run(args: list[str], *, env: dict[str, str] | None = None, cwd: Path | None = None) -> subprocess.CompletedProcess[str]:
+    merged = os.environ.copy()
+    merged["HYPERLEX_OFFLINE"] = "1"
+    merged["HYPERLEX_NO_RATE_LIMIT"] = "1"
+    if env:
+        merged.update(env)
+    return subprocess.run(
+        args,
+        capture_output=True,
+        text=True,
+        cwd=str(cwd or ROOT),
+        env=merged,
+    )
+
+
+def test_hermes_dry_run_unchanged(tmp_path: Path) -> None:
+    home = tmp_path / "home"
+    home.mkdir()
+    env = {
+        "HOME": str(home),
+        "HERMES_HOME": str(home / "profile"),
+    }
+    result = _run(["bash", str(INSTALL), "--dry-run"], env=env)
+    assert result.returncode == 0, result.stderr + result.stdout
+    assert "would staged-validate then activate Hyperlex" in result.stdout
+    assert "Claude personal skill" not in result.stdout
+    assert "Claude plugin dir" not in result.stdout
+    assert not (home / ".claude").exists()
+    assert not (home / "profile" / "skills").exists()
+
+
+def test_claude_dry_run_lists_personal_path(tmp_path: Path) -> None:
+    home = tmp_path / "home"
+    home.mkdir()
+    env = {
+        "HOME": str(home),
+        "HERMES_HOME": str(home / "profile"),
+    }
+    result = _run(["bash", str(INSTALL), "--claude", "--dry-run"], env=env)
+    assert result.returncode == 0, result.stderr + result.stdout
+    assert "would staged-validate then activate Hyperlex" in result.stdout
+    assert "Claude personal skill" in result.stdout
+    assert str(home / ".claude" / "skills" / "hyperlex") in result.stdout
+    assert "hyperlex-demo" in result.stdout
+    assert "hyperlex-settle" in result.stdout
+    assert not (home / ".claude").exists()
+
+
+def test_claude_plugin_dry_run_lists_plugin_path(tmp_path: Path) -> None:
+    home = tmp_path / "home"
+    home.mkdir()
+    env = {
+        "HOME": str(home),
+        "HERMES_HOME": str(home / "profile"),
+    }
+    result = _run(["bash", str(INSTALL), "--claude-plugin", "--dry-run"], env=env)
+    assert result.returncode == 0, result.stderr + result.stdout
+    assert "Claude plugin dir" in result.stdout
+    assert str(home / ".claude" / "plugins" / "hyperlex") in result.stdout
+    assert not (home / ".claude").exists()
+
+
+def test_install_claude_skip_smoke_writes_personal_and_keeps_hermes(tmp_path: Path) -> None:
+    home = tmp_path / "home"
+    home.mkdir()
+    profile = home / "profile"
+    env = {
+        "HOME": str(home),
+        "HERMES_HOME": str(profile),
+    }
+    result = _run(["bash", str(INSTALL), "--claude", "--skip-smoke"], env=env)
+    assert result.returncode == 0, result.stderr + result.stdout
+    hermes = profile / "skills" / "hyperlex"
+    claude = home / ".claude" / "skills" / "hyperlex"
+    assert (hermes / "SKILL.md").is_file()
+    assert (hermes / "scripts" / "hyperlex.py").is_file()
+    assert (claude / "SKILL.md").is_file()
+    assert (claude / "scripts" / "hyperlex.py").is_file()
+    assert (claude / "scripts" / "claude_hlx.sh").is_file()
+    assert (home / ".claude" / "skills" / "hyperlex-demo" / "SKILL.md").is_file()
+    assert (home / ".claude" / "skills" / "hyperlex-settle" / "SKILL.md").is_file()
+    assert not (home / ".hermes").exists()
+
+
+def test_doctor_reports_claude_missing(tmp_path: Path) -> None:
+    home = tmp_path / "empty-home"
+    home.mkdir()
+    env = {
+        "HOME": str(home),
+        "HYPERLEX_SKILL_DIR": "",
+        "CLAUDE_SKILL_DIR": "",
+        "HERMES_SKILL_DIR": "",
+    }
+    # Drop leftover skill-dir hints from the parent environment.
+    result = _run([sys.executable, str(SCRIPT), "doctor"], env=env)
+    assert result.returncode == 0, result.stderr + result.stdout
+    body = json.loads(result.stdout)
+    assert body["ok"] is True
+    assert body["n_failed"] == 0
+    claude = next(c for c in body["checks"] if c["name"] == "claude_host")
+    assert claude["ok"] is True
+    assert claude["message"].startswith("CLAUDE_MISSING")
+
+
+def test_doctor_reports_claude_ok(tmp_path: Path) -> None:
+    home = tmp_path / "home"
+    skill = home / ".claude" / "skills" / "hyperlex"
+    skill.mkdir(parents=True)
+    (skill / "SKILL.md").write_text("# hyperlex\n", encoding="utf-8")
+    (skill / "scripts").mkdir()
+    (skill / "scripts" / "hyperlex.py").write_text("# stub\n", encoding="utf-8")
+    env = {
+        "HOME": str(home),
+        "HYPERLEX_SKILL_DIR": "",
+        "CLAUDE_SKILL_DIR": "",
+        "HERMES_SKILL_DIR": "",
+    }
+    result = _run([sys.executable, str(SCRIPT), "doctor"], env=env)
+    assert result.returncode == 0, result.stderr + result.stdout
+    body = json.loads(result.stdout)
+    assert body["ok"] is True
+    claude = next(c for c in body["checks"] if c["name"] == "claude_host")
+    assert claude["ok"] is True
+    assert claude["message"].startswith("CLAUDE_OK")
+    assert "personal=" in claude["message"]
+
+
+def test_claude_hlx_wrapper_check() -> None:
+    result = _run(["bash", str(ROOT / "scripts" / "claude_hlx.sh"), "check"])
+    assert result.returncode == 0, result.stderr + result.stdout
+    body = json.loads(result.stdout)
+    assert body["ok"] is True
+
+
+def test_plugin_manifest_matches_version() -> None:
+    version = (ROOT / "VERSION").read_text(encoding="utf-8").strip()
+    manifest = json.loads((ROOT / ".claude-plugin" / "plugin.json").read_text(encoding="utf-8"))
+    assert manifest["name"] == "hyperlex"
+    assert manifest["version"] == version
+    assert "Applied Alchemy Labs" in manifest["author"]["name"]
+    assert "github.com/Zero-State-LLC/Hyperlex" in manifest["homepage"]
+
+
+def test_helper_skills_are_short_and_fail_closed() -> None:
+    helpers = (
+        "hyperlex-demo",
+        "hyperlex-wizard",
+        "hyperlex-scan",
+        "hyperlex-analyze",
+        "hyperlex-pending",
+        "hyperlex-settle",
+    )
+    for name in helpers:
+        skill = ROOT / ".claude" / "skills" / name / "SKILL.md"
+        command = ROOT / "commands" / f"{name}.md"
+        body = skill.read_text(encoding="utf-8")
+        assert skill.is_file()
+        assert command.is_file()
+        assert "Never invent" in body or "Do not invent" in body or "do not invent" in body.lower()
+        assert "$HLX" in body
+        assert len(body.splitlines()) < 50
diff --git a/tests/test_doctor.py b/tests/test_doctor.py
index aeb4730..983c55b 100644
--- a/tests/test_doctor.py
+++ b/tests/test_doctor.py
@@ -34,3 +34,7 @@ def test_doctor_ok() -> None:
     assert "brier_null" in names
     assert "api_v1" in names
     assert "golden_corpus" in names
+    assert "claude_host" in names
+    claude = next(c for c in body["checks"] if c["name"] == "claude_host")
+    assert claude["ok"] is True
+    assert claude["message"].startswith("CLAUDE_")