Add portable EVOLVE ChatGPT plugin

.codex-plugin/plugin.json

@@ -1,6 +1,6 @@
 {
   "name": "evolvable-software",
-  "version": "0.6.0",
+  "version": "0.6.1",
   "description": "Assess how ready a product is to evolve itself safely, and how much of that AI drives.",
   "author": {
     "name": "Vellanki Sriharsha",
@@ -21,11 +21,10 @@
   "interface": {
     "displayName": "EVOLVE",
     "shortDescription": "Score self-evolution readiness",
-    "longDescription": "EVOLVE assesses how ready a software product is to evolve itself safely: to take requests through to release, fix the issues users face and find adjacent opportunities, increasingly with AI doing the work. It reports an autonomy level for each loop, how much of it AI drives, the critical controls that fail, exactly what blocks the next level and a fix-first plan, with every score tied to file-level evidence.",
+    "longDescription": "EVOLVE assesses how ready a software product is to evolve itself safely: to take requests through to release, fix the issues users face and find adjacent opportunities, increasingly with AI doing the work. It reports an autonomy level for each loop, how much of it AI drives, the critical controls that fail, exactly what blocks the next level and a fix-first plan, with every score tied to file-level evidence. Runs as a skills-only workflow using supplied repository files and available host tools. Full scoring requires Python execution; without it, EVOLVE provides an evidence review and scoring handoff without inventing computed levels. It does not connect a repository or modify the assessed product.",
     "developerName": "Vellanki Sriharsha",
     "category": "Coding",
     "capabilities": [
-      "Interactive",
       "Read",
       "Write"
     ],
@@ -35,8 +34,8 @@
       "What blocks this product from the next autonomy level?",
       "How much of this product's change loop does AI drive?"
     ],
-    "logo": "./skills/evolvable-software/.claude-plugin/icon.png",
-    "composerIcon": "./skills/evolvable-software/.claude-plugin/icon.png",
+    "logo": "./assets/evolve-icon.png",
+    "composerIcon": "./assets/evolve-icon.png",
     "brandColor": "#2DD4BF"
   }
 }

CHANGELOG.md

@@ -1,5 +1,16 @@
 # Changelog
 
+## EVOLVE v0.6.1, 2026-10-03
+
+### Added
+
+- **ChatGPT plugin packaging.** A root `plugin.json` in the Agent Plugins 1.0 format carries the OpenAI listing fields; the Codex manifest is updated to match. Publication notes and outstanding listing facts are in `docs/publishing/chatgpt.md`.
+
+### Changed
+
+- **Honest fallback when Python is unavailable.** The skill checks whether the host can run the bundled scorer. If it cannot, it gives a clearly unscored evidence review and a hand-off for running the scorer elsewhere, and never computes levels by hand.
+- **Source access.** The repository named in the assessment request counts as authorised to read; the skill still asks before opening any other remote source.
+
 ## EVOLVE v0.6.0, 2026-10-02
 
 ### Added

docs/publishing/chatgpt.md

@@ -0,0 +1,13 @@
+# EVOLVE ChatGPT publication preparation
+
+Implemented: portable Agent Plugins 1.0 root manifest, original v0.6 skill and standard-library Python scorer, existing icon, MIT attribution, honest host capability fallback. No remote MCP or app dependencies. Skills-only submission requires no MCP review cases, demo recording or reviewer credentials.
+
+Draft listing: EVOLVE — Score self-evolution readiness. Author source: Vellanki Sriharsha; selected verified publisher identity still needs confirmation. Website: public source repository. Public release is pending; private creation does not publish a directory listing.
+
+Outstanding facts: verified publisher identity; supported countries (explicit all available countries is an option); whether payments or purchases are involved; public support page/contact arrangement; actual data practices and published privacy policy; published terms.
+
+Do not invent privacy claims from the absence of a server: the host processes user-provided evidence, and any support or analytics practices must be confirmed. Do not equate the MIT source license with service terms.
+
+Once these facts and pages are supplied, populate supportURL, privacyPolicyURL, termsOfServiceURL, review commerce fields and publication countries. Verify all four listing URLs by inspecting their contents. Upload the rebuilt ZIP in the public developer submission portal, confirm imported metadata, complete developer verification and legal/policy attestations, submit for review, and publish the approved release separately.
+
+Example prompts: assess an uploaded repository; explain blockers to the next autonomy level; review how far AI drives the change loop. Repository evidence does not establish production readiness. If Python is unavailable, the expected result is an explicitly unscored evidence review.

plugin.json

@@ -0,0 +1,48 @@
+{
+  "name": "evolvable-software",
+  "version": "0.6.1",
+  "description": "Assess how ready a product is to evolve itself safely, and how much of that AI drives.",
+  "author": {
+    "name": "Vellanki Sriharsha",
+    "url": "https://github.com/harsh51191"
+  },
+  "homepage": "https://github.com/harsh51191/evolvable-software",
+  "repository": "https://github.com/harsh51191/evolvable-software",
+  "license": "MIT",
+  "keywords": [
+    "ai-agents",
+    "self-improving-software",
+    "software-architecture",
+    "governance",
+    "maturity-model",
+    "assessment"
+  ],
+  "$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
+  "extensions": {
+    "com.openai": {
+      "interface": {
+        "displayName": "EVOLVE",
+        "shortDescription": "Score self-evolution readiness",
+        "longDescription": "EVOLVE assesses how ready a software product is to evolve itself safely: to take requests through to release, fix the issues users face and find adjacent opportunities, increasingly with AI doing the work. It reports an autonomy level for each loop, how much of it AI drives, the critical controls that fail, exactly what blocks the next level and a fix-first plan, with every score tied to file-level evidence. Runs as a skills-only workflow using supplied repository files and available host tools. Full scoring requires Python execution; without it, EVOLVE provides an evidence review and scoring handoff without inventing computed levels. It does not connect a repository or modify the assessed product.",
+        "developerName": "Vellanki Sriharsha",
+        "category": "Coding",
+        "capabilities": [
+          "Read",
+          "Write"
+        ],
+        "websiteURL": "https://github.com/harsh51191/evolvable-software",
+        "defaultPrompt": [
+          "Assess how ready this repository is to evolve itself safely",
+          "What blocks this product from the next autonomy level?",
+          "How much of this product's change loop does AI drive?"
+        ],
+        "logo": "./assets/evolve-icon.png",
+        "composerIcon": "./assets/evolve-icon.png",
+        "brandColor": "#2DD4BF"
+      },
+      "publication": {
+        "release_notes": "First portable ChatGPT plugin package. Includes EVOLVE v0.6 rubric, deterministic Python scorer, evidence workflow, remediation, existing branding, and host capability checks."
+      }
+    }
+  }
+}

skills/evolvable-software/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "name": "evolvable-software",
   "displayName": "EVOLVE: Self-Evolution Readiness",
-  "version": "0.6.0",
+  "version": "0.6.1",
   "description": "Assess how ready a software product is to evolve itself safely, and how much of that AI drives: autonomy levels for request-to-release, issue-to-fix and opportunity-to-expansion, critical controls, what blocks the next level, and a fix-first plan, all from code evidence.",
   "author": {
     "name": "Vellanki Sriharsha",
@@ -12,5 +12,13 @@
   "documentationUrl": "https://github.com/harsh51191/evolvable-software#readme",
   "supportUrl": "https://github.com/harsh51191/evolvable-software/issues",
   "license": "MIT",
-  "keywords": ["ai-agents", "self-improving-software", "software-architecture", "governance", "maturity-model", "ai-readiness", "assessment"]
+  "keywords": [
+    "ai-agents",
+    "self-improving-software",
+    "software-architecture",
+    "governance",
+    "maturity-model",
+    "ai-readiness",
+    "assessment"
+  ]
 }

skills/evolvable-software/SKILL.md

@@ -55,6 +55,12 @@ The result is a reading for one product, not a leaderboard. Different archetypes
 - Roadmap scoring or future-state promises.
 - Declaring production readiness from repository evidence alone.
 
+## Host capability check
+
+Use the repository URL or files the user supplied as authorization to inspect that source with available tools. Do not imply this plugin adds GitHub access, a hosted server or a Python runtime. Uploaded repository archives and user-supplied evidence also work. If sources cannot be accessed, request a repository archive or relevant files and record the evidence gap.
+
+Before scoring, confirm that this host can execute Python and access the bundled scripts and references. If so, use the bundled scorer unchanged. If not, gather evidence and explain the rubric, but do not calculate Software Autonomy Levels or AI Readiness by hand. Supply a clearly marked unscored evidence report and an input/command handoff for running the scorer in a Python-capable environment. Never claim a script ran or an assessment passed without actual output.
+
 ## Procedure
 
 ### 1. Establish scope
@@ -65,7 +71,7 @@ Scope must include every first-party repository that implements the product's co
 
 Read `references/archetypes.md` before choosing among `configurable-application-platform`, `agent-runtime`, `developer-platform`, `focused-application` and `other`.
 
-Assess only repositories and documents the user supplied or authorised. Never mutate the target repository. Ask before cloning or accessing a new remote source.
+Assess only repositories and documents the user supplied or authorised. Never mutate the target repository. The supplied target is authorized by the assessment request; ask before accessing additional remote sources outside that scope.
 
 ### 2. Create the input