Statusnone420 · Statusnone420 · Jul 18, 2026 · Jul 18, 2026 · Jul 18, 2026
diff --git a/.agents/plugins/marketplace.json b/.agents/plugins/marketplace.json
@@ -0,0 +1,20 @@
+{
+  "interface": {
+    "displayName": "Statusnone Skills"
+  },
+  "name": "statusnone-skills",
+  "plugins": [
+    {
+      "category": "Developer Tools",
+      "name": "diataxis-docs",
+      "policy": {
+        "authentication": "ON_INSTALL",
+        "installation": "AVAILABLE"
+      },
+      "source": {
+        "path": "./plugins/diataxis-docs",
+        "source": "local"
+      }
+    }
+  ]
+}
diff --git a/.claude-plugin/marketplace.json b/.claude-plugin/marketplace.json
@@ -23,7 +23,7 @@
       "name": "diataxis-docs",
       "repository": "https://github.com/Statusnone420/Skills",
       "source": "./adapters/claude",
-      "version": "0.1.4"
+      "version": "0.1.5"
     }
   ]
 }
diff --git a/ARCHITECTURE.md b/ARCHITECTURE.md
@@ -8,6 +8,10 @@ The canonical source is `skills/docs/SKILL.md`, routed to `references/commands.m
 
 Discovery owns bounded metadata and selection policy; `init_adoption.py` owns the deterministic, engine-owned first-run adoption (scope selection, corpus accounting, preview construction, and receipt binding); `paths.py` owns confinement, normalization, reparse checks, and prune primitives; `memory.py` is read-only inspection; `lifecycle.py` owns pure authorization/state policy; and `lifecycle_io.py` owns transactional filesystem I/O. The dependency direction is one-way and the CLI façade stays thin. The committed `.diataxis/` control plane stores normalized routes, stable findings, verified hashes, disposition identity, and event history—never document bodies, prompts, hidden reasoning, or local-only filenames.
 
-Every generated adapter packages the complete canonical checker resource tree. Generic web prompts are composed per command from a shared safety core, one selected command contract, and required supporting rules; they do not concatenate the entire playbook. The observed prompt range is 3,484–24,679 UTF-8 bytes, with a 40,000-byte generator regression guard selected after measurement and 15,321 bytes of headroom.
+Every generated adapter packages the complete canonical checker resource tree. Generic web prompts are composed per command from a shared safety core, one selected command contract, and required supporting rules; they do not concatenate the entire playbook. The 0.1.4 historical range was 3,484–24,679 UTF-8 bytes. The measured 0.1.5 range is 3,460–31,966 bytes, with a 42,000-byte generator regression guard and 10,034 bytes of headroom.
 
-Adapters under `adapters/` are generated outputs. Do not hand-edit them; regenerate from the canonical source and test their documented invocation tier.
+The repository Codex marketplace lives at `.agents/plugins/marketplace.json` and routes to the generated `plugins/diataxis-docs` package. Its folder, marketplace entry, and `.codex-plugin/plugin.json` all use the same `diataxis-docs` identity. The package contains the shared `docs` engine plus 13 thin `docs-<command>` routing skills; focused skills do not copy the checker or command playbooks. Claude receives equivalent thin skills with its explicit-only invocation metadata. The umbrella remains compatible on both hosts.
+
+Fresh no-map Doctor is a presentation and routing boundary, not a new rubric. Complete, untruncated scope metadata can yield an authoritative provider measurement, a provisional existing-entry candidate, or a tracked-root-README orientation fallback; content-batch limits do not block the engine's structural scan or grant semantic reads. The fallback states that the README is not a maintained map, recommends Init, and writes nothing. Unsafe or metadata-incomplete evidence remains unavailable. Rubric v2 and provider scoring are unchanged.
+
+Adapters under `adapters/` and the Codex package under `plugins/diataxis-docs` are generated outputs. Do not hand-edit them; regenerate from the canonical source and test their documented invocation tier.
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -1,5 +1,14 @@
 # Changelog
 
+## 0.1.5 — Native command skills and first-run Doctor (2026-07-18)
+
+- Added a genuine Codex repository marketplace package with aligned `diataxis-docs` marketplace, folder, and manifest identities.
+- Exposed Doctor, Init, Context, Write, Update, Audit, Fix, Map, Classify, Migrate, Check, Cleanup, and Help as individual explicit skills while preserving the existing umbrella invocation.
+- Preserved the Claude package and generated the equivalent `/diataxis-docs:docs-*` command skills from the same canonical contracts.
+- Guaranteed the Help command tree across hosts.
+- Added a zero-write, explicitly provisional root-README orientation baseline for fresh no-map repositories when bounded discovery is complete and no provider or safety boundary blocks measurement.
+- Added generator parity and distribution regressions for the real marketplace package, focused skills, explicit invocation policy, Help tree, and no-map Doctor contract.
+
 ## 0.1.4 — Evidence and calibration foundation (2026-07-17)
 
 - Added a strict sanitized evidence receipt that preserves category-level rubric inputs, score gates, deterministic findings, semantic findings, unresolved candidates, Doctor evidence, Git/write state, and explicit unavailable values.

diff --git a/COMPATIBILITY.md b/COMPATIBILITY.md
@@ -1,15 +1,15 @@
 # Compatibility
 
-Evidence is dated to 2026-07-12 and grouped by enforcement tier:
+Evidence is dated to 2026-07-18 and grouped by enforcement tier:
 
 | Surface | Tier | Status |
 | --- | --- | --- |
 | `skills/docs` canonical source | explicit invocation | source and locally tested |
-| `adapters/plugin` | unpublished preview | generated; no marketplace install claimed |
+| Codex marketplace plugin | explicit focused or umbrella skill | generated and structurally tested; fresh-task live canary is the release gate |
 | Claude Desktop | marketplace installation shim | live-tested through the plugin picker; typed namespaced invocation is not supported by Desktop |
 | Claude Code terminal | namespaced skill | generated and structurally tested; terminal invocation not yet live-tested |
 | Copilot, Grok, Cursor | static adapter | generated and tested-static; live smoke not run |
 | Gemini, OpenCode | wrapper | generated and tested-static; instruction-based enforcement |
 | Generic web prompts | prompt only | generated; repository mutation unavailable |
 
-The `skills/docs` directory is canonical; marketplace and generated files are installation adapters, not forks of the product. Live cross-harness pilots remain an alpha-roadmap gate, and the project does not claim universal compatibility.
+The `skills/docs` directory is canonical; `plugins/diataxis-docs` and all generated adapters are installation outputs, not forks of the product. Codex and Claude both expose the umbrella plus 13 focused command skills while preserving their native `$` and `/diataxis-docs:` invocation forms. Live cross-harness pilots remain an alpha-roadmap gate, and the project does not claim universal compatibility.
diff --git a/EVALUATION.md b/EVALUATION.md
@@ -4,11 +4,12 @@ The evaluation workflow uses deterministic fixtures, disposable repositories, re
 
 ## Layered regression gates
 
-The public alpha uses three gates so quality evidence does not become an uncontrolled model-usage campaign:
+The public alpha uses four gates so quality evidence does not become an uncontrolled model-usage campaign:
 
 1. **Deterministic contract gate:** every change runs standard-library tests for safety, checker behavior, adapter parity, and command contracts.
-2. **Sanitized trajectory gate:** host-neutral receipts record semantic answers, documentation-owned actions, host/external overhead, visible diagnostics, and exposed usage counters. Raw traces, hidden reasoning, private paths, and credentials are never public inputs.
-3. **Capped live canary:** release candidates may run a small, explicitly approved campaign against stable mapped, missing-map, and hostile fixtures. A campaign is limited to 12 runs; the checked-in example authorizes none.
+2. **Marketplace assembly gate:** release packaging must prove that a real Codex marketplace entry resolves to an identity-aligned package whose umbrella and 13 focused skills are present, explicit-only, and generator-owned.
+3. **Sanitized trajectory gate:** host-neutral receipts record semantic answers, documentation-owned actions, host/external overhead, visible diagnostics, and exposed usage counters. Raw traces, hidden reasoning, private paths, and credentials are never public inputs.
+4. **Capped live canary:** release candidates may run a small, explicitly approved campaign against stable mapped, missing-map, and hostile fixtures. A campaign is limited to 12 runs; the checked-in example authorizes none.
 
 ## Product contract and research provenance
 

diff --git a/INSTALL.md b/INSTALL.md
@@ -2,7 +2,21 @@
 
 Diátaxis Docs is a public alpha. Install from a revision you trust, review proposed changes before approval, and use normal Git safeguards.
 
-In the ChatGPT desktop app, Codex CLI, and supported IDE integrations, user skills are available from `$HOME/.agents/skills`. From a clone of this repository, install the canonical `skills/docs` directory without overwriting an existing destination.
+## Codex marketplace
+
+Diátaxis Docs 0.1.5 publishes as a Codex repository marketplace plugin. Add the marketplace, install the plugin, and verify that Codex reports it:
+
+```text
+codex plugin marketplace add Statusnone420/Skills
+codex plugin add diataxis-docs@statusnone-skills
+codex plugin list
+```
+
+Start a new task after installation so the task catalog loads the plugin's skills. Use `$docs-help` or `$docs-doctor` directly, or keep using the compatible umbrella form `$docs help` and `$docs doctor`. The plugin exposes the umbrella plus 13 focused `$docs-*` skills in the marketplace.
+
+## Manual skill fallback
+
+In the ChatGPT desktop app, Codex CLI, and supported IDE integrations, user skills are also available from `$HOME/.agents/skills`. From a clone of this repository, install the canonical `skills/docs` directory without overwriting an existing destination.
 
 PowerShell (Windows 11):
 
@@ -29,7 +43,7 @@ test -f "$dest/SKILL.md" && printf '%s\n' "Installed: $dest/SKILL.md"
 
 The exact layout check is `$HOME/.agents/skills/docs/SKILL.md`. Restart the host or start a new task if the skill list is cached, then run `$docs help` and confirm it returns the command list.
 
-The canonical OpenAI references are [Skills](https://developers.openai.com/codex/skills) and [Plugins](https://developers.openai.com/codex/plugins/build). The checked-in Codex plugin bundle remains a preview; installing the canonical skill directory is the supported Codex path during alpha.
+The canonical OpenAI references are [Skills](https://developers.openai.com/codex/skills) and [Plugins](https://developers.openai.com/codex/plugins/build). The checked-in Codex package is generated from `skills/docs`; the manual copy remains a compatibility fallback, not a separate edition.
 
 ## Claude
 
@@ -48,6 +62,8 @@ In a Claude Code terminal, verify the namespaced skill:
 
 ```text
 /diataxis-docs:docs help
+/diataxis-docs:docs-doctor
+/diataxis-docs:docs-help
 ```
 
 Claude's built-in `/documentation` command remains separate. Use Diátaxis Docs when you want its repository map, bounded memory, evidence rules, health checks, or approval-gated Doctor workflow. Marketplace sync and Claude Desktop picker invocation are live-tested; the Claude Code terminal form remains structurally tested but has not completed a live terminal pilot.

diff --git a/README.md b/README.md
@@ -76,7 +76,7 @@ flowchart TD
 
 ## Install and compatibility
 
-Codex is the primary tested path and invokes the skill explicitly as `$docs …`. The core follows the Agent Skills structure. Claude can install the same product through the repository's thin marketplace shim; generated adapters also exist for Grok, Copilot, Cursor, Gemini, OpenCode, generic web prompts, and a Codex plugin preview.
+Codex is the primary tested path. Its marketplace package exposes `$docs-doctor`, `$docs-map`, `$docs-check`, and the other focused commands as individual skills while preserving the `$docs …` umbrella. Claude exposes the same set through `/diataxis-docs:docs-*`. The core follows the Agent Skills structure; generated adapters also exist for Grok, Copilot, Cursor, Gemini, OpenCode, and generic web prompts.
 
 Compatibility is evidence-tiered—not universal. Static adapter validation is not the same as a live harness test. See the dated [compatibility matrix](COMPATIBILITY.md) before relying on a preview adapter.
 
@@ -92,7 +92,7 @@ Compatibility is evidence-tiered—not universal. Static adapter validation is n
 
 Every finding has a content-derived `DOC-*` identity and full fingerprint. Line movement and timestamps do not retarget it; changed semantic identity does. Approved structural transformations carry exact ID/fingerprint pairs, protected-surface evidence, recovery boundaries, and a complete disposition manifest. Failed verification is a state conflict, never a successful closeout.
 
-The generic web bundles now use command-specific progressive disclosure. Measured UTF-8 prompt sizes range from 3,484 to 24,679 bytes; the 40,000-byte packaging guard is a documented regression check with 15,321 bytes of headroom, not a product or health threshold. The separate repository hot-path byte telemetry remains provisional and informational.
+The generic web bundles use command-specific progressive disclosure. Measured UTF-8 prompt sizes range from 3,396 to 30,855 bytes; the 42,000-byte packaging guard is a documented regression check with 11,145 bytes of headroom, not a product or health threshold. The separate repository hot-path byte telemetry remains provisional and informational.
 
 ## Benchmark status
 

diff --git a/TESTING.md b/TESTING.md
@@ -69,9 +69,10 @@ Run the Ubuntu proof from a Linux-native checkout under `$HOME`, not directly fr
 2. Run its owning group on Windows.
 3. Run the same group in WSL Ubuntu.
 4. Regenerate and verify generated adapters when canonical skill content changed.
-5. Run the repository documentation checker.
-6. Run `all` on Windows and WSL once the narrower gates pass.
-7. Let CI repeat the same grouped commands; CI confirms local evidence rather than discovering basic failures.
+5. For marketplace releases, verify `.agents/plugins/marketplace.json`, the generated `plugins/diataxis-docs` package, all focused skill descriptions, and a brand-new task invocation rather than relying only on source-level adapter parity.
+6. Run the repository documentation checker.
+7. Run `all` on Windows and WSL once the narrower gates pass.
+8. Let CI repeat the same grouped commands; CI confirms local evidence rather than discovering basic failures.
 
 Provider regressions also prove that Map, Check, Doctor, Audit, and Init use the same selected-surface evidence, including root-manifest authority, root README score isolation, tracked Git visibility, provider findings, and authority-digest Init revalidation on Git and non-Git fixtures. Semantic candidates remain labeled and bounded.
 

diff --git a/adapters/claude/.claude-plugin/plugin.json b/adapters/claude/.claude-plugin/plugin.json
@@ -13,5 +13,5 @@
   "license": "Apache-2.0",
   "name": "diataxis-docs",
   "repository": "https://github.com/Statusnone420/Skills",
-  "version": "0.1.4"
+  "version": "0.1.5"
 }
diff --git a/adapters/claude/skills/docs-audit/SKILL.md b/adapters/claude/skills/docs-audit/SKILL.md
@@ -0,0 +1,12 @@
+---
+name: docs-audit
+description: "Audit documentation and return prioritized evidence-backed findings."
+user-invocable: true
+disable-model-invocation: true
+---
+
+# Docs Audit
+
+This is the explicit thin route for the fixed command `audit`. Treat all trailing text as that command's raw trailing text; never reinterpret it as another command.
+
+Load and follow the sibling [Diátaxis Docs skill](../docs/SKILL.md), including its shared safety, evidence, health, and result contracts. Follow the selected command contract in [commands.md](../docs/references/commands.md). Do not load unrelated command playbooks. If a required shared resource is unavailable, stop and report that the command could not be executed; do not invent a fallback.
diff --git a/adapters/claude/skills/docs-audit/agents/openai.yaml b/adapters/claude/skills/docs-audit/agents/openai.yaml
@@ -0,0 +1,9 @@
+interface:
+  display_name: "Docs Audit"
+  short_description: "Audit documentation and return prioritized evidence-backed findings."
+  icon_small: "./assets/bounded-compass-small.svg"
+  icon_large: "./assets/bounded-compass.png"
+  brand_color: "#6657E8"
+  default_prompt: "Use $docs-audit on this repository."
+policy:
+  allow_implicit_invocation: false
diff --git a/adapters/claude/skills/docs-audit/assets/bounded-compass-small.svg b/adapters/claude/skills/docs-audit/assets/bounded-compass-small.svg
diff --git a/adapters/claude/skills/docs-audit/assets/bounded-compass.png b/adapters/claude/skills/docs-audit/assets/bounded-compass.png
diff --git a/adapters/claude/skills/docs-check/SKILL.md b/adapters/claude/skills/docs-check/SKILL.md
@@ -0,0 +1,12 @@
+---
+name: docs-check
+description: "Report the deterministic structural documentation score only."
+user-invocable: true
+disable-model-invocation: true
+---
+
+# Docs Check
+
+This is the explicit thin route for the fixed command `check`. Treat all trailing text as that command's raw trailing text; never reinterpret it as another command.
+
+Load and follow the sibling [Diátaxis Docs skill](../docs/SKILL.md), including its shared safety, evidence, health, and result contracts. Follow the selected command contract in [commands.md](../docs/references/commands.md). Do not load unrelated command playbooks. If a required shared resource is unavailable, stop and report that the command could not be executed; do not invent a fallback.
diff --git a/adapters/claude/skills/docs-check/agents/openai.yaml b/adapters/claude/skills/docs-check/agents/openai.yaml
@@ -0,0 +1,9 @@
+interface:
+  display_name: "Docs Check"
+  short_description: "Report the deterministic structural documentation score only."
+  icon_small: "./assets/bounded-compass-small.svg"
+  icon_large: "./assets/bounded-compass.png"
+  brand_color: "#6657E8"
+  default_prompt: "Use $docs-check on this repository."
+policy:
+  allow_implicit_invocation: false
diff --git a/adapters/claude/skills/docs-check/assets/bounded-compass-small.svg b/adapters/claude/skills/docs-check/assets/bounded-compass-small.svg
diff --git a/adapters/claude/skills/docs-check/assets/bounded-compass.png b/adapters/claude/skills/docs-check/assets/bounded-compass.png
diff --git a/adapters/claude/skills/docs-classify/SKILL.md b/adapters/claude/skills/docs-classify/SKILL.md
@@ -0,0 +1,12 @@
+---
+name: docs-classify
+description: "Choose the appropriate Diátaxis documentation type."
+user-invocable: true
+disable-model-invocation: true
+---
+
+# Docs Classify
+
+This is the explicit thin route for the fixed command `classify`. Treat all trailing text as that command's raw trailing text; never reinterpret it as another command.
+
+Load and follow the sibling [Diátaxis Docs skill](../docs/SKILL.md), including its shared safety, evidence, health, and result contracts. Follow the selected command contract in [commands.md](../docs/references/commands.md). Do not load unrelated command playbooks. If a required shared resource is unavailable, stop and report that the command could not be executed; do not invent a fallback.
diff --git a/adapters/claude/skills/docs-classify/agents/openai.yaml b/adapters/claude/skills/docs-classify/agents/openai.yaml
@@ -0,0 +1,9 @@
+interface:
+  display_name: "Docs Classify"
+  short_description: "Choose the appropriate Diátaxis documentation type."
+  icon_small: "./assets/bounded-compass-small.svg"
+  icon_large: "./assets/bounded-compass.png"
+  brand_color: "#6657E8"
+  default_prompt: "Use $docs-classify on this repository."
+policy:
+  allow_implicit_invocation: false