project-blueprint-system 0.2.0-demo.0

package/.agents/plugins/marketplace.json

@@ -0,0 +1,20 @@
+{
+  "name": "project-blueprint-demo",
+  "interface": {
+    "displayName": "Project Blueprint Demo Skills"
+  },
+  "plugins": [
+    {
+      "name": "project-blueprint-system",
+      "source": {
+        "source": "local",
+        "path": "./plugins/project-blueprint-system"
+      },
+      "policy": {
+        "installation": "AVAILABLE",
+        "authentication": "ON_INSTALL"
+      },
+      "category": "Productivity"
+    }
+  ]
+}

package/.claude-plugin/marketplace.json

@@ -0,0 +1,26 @@
+{
+  "name": "project-blueprint-demo",
+  "owner": {
+    "name": "Mahmoud Eyada"
+  },
+  "plugins": [
+    {
+      "name": "project-blueprint-system",
+      "source": "./plugins/project-blueprint-system",
+      "description": "Operate a software project from client intake through delivery and operations inside the agent conversation.",
+      "version": "0.2.0-demo.0",
+      "author": {
+        "name": "Mahmoud Eyada"
+      },
+      "homepage": "https://github.com/Shiro7/project-blueprint-system",
+      "repository": "https://github.com/Shiro7/project-blueprint-system",
+      "license": "UNLICENSED",
+      "keywords": [
+        "project-blueprint",
+        "solution-architecture",
+        "agent-skill"
+      ],
+      "category": "Productivity"
+    }
+  ]
+}

package/DEMO_STATUS.md

@@ -0,0 +1,37 @@
+# Demo Status
+
+This is an early public evaluation release.
+
+## What testers should evaluate
+
+- Does L1 create the minimum useful project blueprint without excessive documentation?
+- Does L2 separate the artifacts that need independent decisions and review?
+- Does L3 expose sufficient depth without forcing every artifact to be applicable?
+- Can an agent correctly distinguish facts, assumptions, decisions, requirements, verification, and evidence?
+- Does information move clearly from discovery into design, architecture, delivery, and testing?
+- Can the project remain synchronized while Figma and code are implemented together?
+- Are skipped artifacts explicitly marked or explained rather than silently ignored?
+- Can a user install and operate the Skill entirely from the agent UI without a terminal?
+- Does the same `SKILL.md` behave consistently in Codex, ChatGPT Work, Claude, and other compatible hosts?
+
+## Known demo gaps
+
+- Artifact-specific question sets need testing and refinement.
+- Domain-level profile overrides are represented in the manifest but are not generated independently yet.
+- Conditional modules currently create activation records and shared guidance; full module-specific artifact packs are planned.
+- Profile upgrade and framework migration workflows are planned.
+- Cross-agent behavior has not been certified.
+- The public ChatGPT/Codex Plugin Directory listing has not been submitted or approved.
+
+## Feedback format
+
+Please include:
+
+1. Agent and version used.
+2. Project type and selected level.
+3. Installation path and invocation used.
+4. Information supplied to the agent.
+5. Files that were useful, missing, duplicated, or unclear.
+6. Decisions the system failed to request.
+7. Verification or evidence gaps.
+8. Suggested profile or module changes.

package/README.md

@@ -0,0 +1,121 @@
+# Project Blueprint System
+
+> **Demo release — not production-ready.** The workflow, profile composition, and cross-agent behavior are being tested and may change before version 1.0.
+
+Project Blueprint System is an installable Agent Skill that carries a software project from the first client conversation through discovery, requirements, UX and Figma, solution and data architecture, security, implementation, verification, release, and operations.
+
+## No Terminal Required
+
+The product is the Skill, not the command line. After installation, use it from the conversation:
+
+> Start a blueprint for this project. Use what I already sent, recommend L1, L2, or L3, and ask me only for the next blocking decision.
+
+The agent then creates or updates the project files directly. If the host cannot write to a workspace, it returns the same blueprint as a downloadable ZIP.
+
+## Download
+
+- [Standalone Agent Skill ZIP](dist/project-blueprint-system-skill-0.2.0-demo.0.zip) — for Claude and other hosts that accept an Agent Skill ZIP.
+- [Cross-agent Plugin ZIP](dist/project-blueprint-system-plugin-0.2.0-demo.0.zip) — contains both Codex/ChatGPT and Claude plugin manifests.
+
+## Install Inside an Agent
+
+### Codex
+
+Open a Codex conversation, invoke `$skill-installer`, and ask it to install:
+
+```text
+Install the Project Blueprint System skill from:
+https://github.com/Shiro7/project-blueprint-system/tree/main/plugins/project-blueprint-system/skills/project-blueprint-system
+```
+
+After installation, select **Project Blueprint System** from Skills or mention `$project-blueprint-system` in the conversation.
+
+### ChatGPT
+
+The same plugin package is prepared for the universal ChatGPT and Codex Plugin Directory. Its public directory listing is not live during this demo stage. In ChatGPT Work environments that expose Skill Installer, invoke `@skill-installer` with the same GitHub skill path above.
+
+Once the plugin is accepted into the public directory, installation becomes a normal **Install** action from ChatGPT or Codex without a package manager.
+
+### Claude
+
+1. Download the [Standalone Agent Skill ZIP](dist/project-blueprint-system-skill-0.2.0-demo.0.zip).
+2. In Claude, open **Customize > Skills**.
+3. Choose **+ > Create skill > Upload a skill**.
+4. Upload the ZIP and enable **Project Blueprint System**.
+
+For Claude Code Desktop, open **Plugins**, add `Shiro7/project-blueprint-system` as a marketplace when the UI offers **Add marketplace**, then install **project-blueprint-system**.
+
+### Other Agents
+
+Import the standalone ZIP or point the agent's skill installer at:
+
+```text
+plugins/project-blueprint-system/skills/project-blueprint-system/
+```
+
+The skill follows the open Agent Skills `SKILL.md` structure. Host-specific installation controls still depend on the agent product.
+
+## What the Skill Does
+
+- Selects one stable execution profile instead of generating uncontrolled documentation.
+- **L1 Rapid:** 17 composite artifacts for a quick but complete path.
+- **L2 Standard:** 51 independently governed artifacts for normal production work.
+- **L3 Comprehensive:** detailed artifacts for complex, sensitive, regulated, or multi-team products.
+- Accepts normal chat messages, documents, Figma links, screenshots, repositories, ERDs, spreadsheets, tests, and logs.
+- Classifies information as sources, facts, assumptions, questions, decisions, requirements, verification, and evidence.
+- Keeps requirements, UX, architecture, security, implementation, tests, release, and operations connected through stable IDs and handoffs.
+- Records unused artifacts as `NOT_APPLICABLE` with a reason, owner, evidence, approval, and reopen condition.
+- Validates structure and content by inspecting the project files; it does not ask the user to run a command.
+
+## Example Conversation
+
+```text
+User: Start a blueprint for this clinic portal. It has patients, doctors,
+admins, appointments, payments, Arabic and English, and a mobile app.
+
+Skill: I recommend L2 Standard with deeper security and privacy coverage.
+Before I create the first project set, I need one blocking decision:
+will the portal store clinical records or only appointment metadata?
+```
+
+Later, the user can simply send another message:
+
+```text
+The client confirmed that doctors can reschedule appointments, but patients
+can only cancel up to 24 hours before the appointment. Here is the Figma link.
+```
+
+The Skill routes the information to the canonical requirements, roles, journeys, Figma, authorization, API, verification, and traceability artifacts, then reports what changed and what decision comes next.
+
+## Repository Layout
+
+```text
+project-blueprint-system/
+├── .agents/plugins/marketplace.json        # Codex/ChatGPT demo marketplace
+├── .claude-plugin/marketplace.json         # Claude demo marketplace
+├── plugins/project-blueprint-system/
+│   ├── .codex-plugin/plugin.json
+│   ├── .claude-plugin/plugin.json
+│   └── skills/project-blueprint-system/
+│       ├── SKILL.md
+│       ├── agents/openai.yaml
+│       ├── assets/
+│       └── references/
+└── dist/
+    ├── project-blueprint-system-skill-0.2.0-demo.0.zip
+    └── project-blueprint-system-plugin-0.2.0-demo.0.zip
+```
+
+The JavaScript package remains only as a compatibility and test harness for this demo. It is not the intended user experience.
+
+## Demo Limitations
+
+- The artifact question banks and generated contracts still need real-world testing.
+- L2 and L3 need evaluation across different project types and agent products.
+- Domain-level profile overrides are defined but still need more cross-agent tests.
+- Conditional modules are preliminary.
+- The system does not guarantee legal, regulatory, privacy, accessibility, or security compliance.
+- A public ChatGPT/Codex Plugin Directory listing requires submission and review.
+- This demo is `UNLICENSED`; evaluation is allowed, but a production and redistribution license has not been selected.
+
+See [DEMO_STATUS.md](DEMO_STATUS.md) for the test checklist and feedback format.

package/bin/project-blueprint.js

@@ -0,0 +1,8 @@
+#!/usr/bin/env node
+
+const { run } = require('../lib/cli');
+
+run(process.argv.slice(2)).catch((error) => {
+  console.error(`project-blueprint: ${error.message}`);
+  process.exitCode = 1;
+});

package/framework/AGENT_OPERATING_CONTRACT.md

@@ -0,0 +1,13 @@
+# Agent Operating Contract
+
+This framework is a demo-stage operating system, not a documentation checklist.
+
+For every information batch:
+
+1. Preserve sources.
+2. Separate facts, assumptions, questions, decisions, requirements, verification, and evidence.
+3. Update the canonical owner artifact.
+4. Propagate stable IDs through downstream design, architecture, implementation, testing, and operations.
+5. Report contradictions, blockers, required decisions, verification, evidence, and next step.
+
+Never silently skip an applicable artifact. Never claim approval or verification without an owner and inspectable basis.