feat: add subagent orchestration kit

.agents/plugins/marketplace.json

@@ -0,0 +1,20 @@
+{
+  "name": "changehack-local-plugins",
+  "interface": {
+    "displayName": "changeHack Local Plugins"
+  },
+  "plugins": [
+    {
+      "name": "subagent-orchestration-kit",
+      "source": {
+        "source": "local",
+        "path": "./plugins/subagent-orchestration-kit"
+      },
+      "policy": {
+        "installation": "AVAILABLE",
+        "authentication": "ON_INSTALL"
+      },
+      "category": "Productivity"
+    }
+  ]
+}

plugins/subagent-orchestration-kit/.codex-plugin/plugin.json

@@ -0,0 +1,45 @@
+{
+  "name": "subagent-orchestration-kit",
+  "version": "0.1.0",
+  "description": "Reusable role prompts and orchestration rules for idea generation, implementation, evaluation, and error checking.",
+  "author": {
+    "name": "Akira Kakiuchi",
+    "email": "support@example.com",
+    "url": "https://github.com/Asheze1127"
+  },
+  "homepage": "https://github.com/Asheze1127/HackHub",
+  "repository": "https://github.com/Asheze1127/HackHub",
+  "license": "UNLICENSED",
+  "keywords": [
+    "subagent",
+    "orchestration",
+    "evaluation",
+    "review",
+    "planning"
+  ],
+  "skills": "./skills/",
+  "interface": {
+    "displayName": "Subagent Orchestration Kit",
+    "shortDescription": "Reusable sub-agent playbooks for product work and implementation loops",
+    "longDescription": "Packages reusable role definitions, project-context handling, and an implementation loop for idea consulting, coding, evaluation, error checking, and orchestration across repos.",
+    "developerName": "Akira Kakiuchi",
+    "category": "Productivity",
+    "capabilities": [
+      "Interactive",
+      "Read",
+      "Write"
+    ],
+    "websiteURL": "https://github.com/Asheze1127/HackHub",
+    "privacyPolicyURL": "https://github.com/Asheze1127/HackHub",
+    "termsOfServiceURL": "https://github.com/Asheze1127/HackHub",
+    "defaultPrompt": [
+      "Use the orchestrator cycle to plan and ship the next slice.",
+      "Use idea-consultant to propose feature options from project docs.",
+      "Use error-checker to compare implementation against specs."
+    ],
+    "brandColor": "#0F766E",
+    "composerIcon": "./assets/subagent-kit.svg",
+    "logo": "./assets/subagent-kit.svg",
+    "screenshots": []
+  }
+}

plugins/subagent-orchestration-kit/README.md

@@ -0,0 +1,60 @@
+# Subagent Orchestration Kit
+
+This plugin packages a reusable improvement loop for product work and implementation work.
+
+It does **not** add new first-class agent types to Codex. Instead, it gives you a portable set of:
+
+- role definitions
+- project-context files
+- orchestration rules
+- usage guidance you can reuse in other repositories
+
+## Included roles
+
+- `idea-consultant`
+  - proposes feature options and product slices from source docs
+- `coding-worker`
+  - implements one bounded slice with clear file ownership
+- `implementation-evaluator`
+  - verifies acceptance criteria and test coverage for the slice
+- `error-checker`
+  - compares code against specs, product direction, and architectural constraints
+- `reviewer`
+  - reviews the concrete diff for regressions, maintainability, and missed tests
+- `orchestrator-cycle`
+  - coordinates the loop and keeps scope under control
+
+## Current repo context
+
+This copy includes a project context file for the current HackHub direction:
+
+- [contexts/hackhub-product-overview.md](./contexts/hackhub-product-overview.md)
+
+That context merges:
+
+- `changeHack`
+- `progress-checker`
+- `hack-evaluater`
+- the user override that the product must support multiple hackathons and serve participants, sponsors, and organizers
+
+## How to use in this repo
+
+1. Start with `orchestrator-cycle`.
+   Use [contexts/hackhub-product-overview.md](./contexts/hackhub-product-overview.md) as the project overview file.
+2. Use `idea-consultant` to decide the smallest valuable slice.
+3. Give `coding-worker` exclusive ownership of the write scope.
+4. Run `error-checker` against docs and code.
+5. Run `implementation-evaluator` against the agreed acceptance criteria.
+6. Repeat in small slices.
+
+## How to reuse in another repo
+
+1. Copy `plugins/subagent-orchestration-kit` into the target repo.
+2. Copy or recreate `.agents/plugins/marketplace.json`, or merge the plugin entry into the target repo marketplace.
+3. Replace `contexts/hackhub-product-overview.md` with a project-specific overview based on [contexts/project-overview-template.md](./contexts/project-overview-template.md).
+4. Update the source-of-truth file list inside the skill docs if the target repo uses different spec paths.
+5. Keep the same role split unless the target repo needs fewer roles.
+
+## Portability note
+
+If the plugin is not installed in a target environment, the files still work as prompt packs and operating docs. You can open the relevant `SKILL.md` files and use them directly when spawning helper agents.

plugins/subagent-orchestration-kit/agents/openai.yaml

@@ -0,0 +1,6 @@
+interface:
+  display_name: "Subagent Orchestration Kit"
+  short_description: "Reusable role prompts for idea consulting, coding, evaluation, and implementation review"
+  icon_small: "./assets/subagent-kit.svg"
+  icon_large: "./assets/subagent-kit.svg"
+  default_prompt: "Use the subagent orchestration kit to plan the next slice, implement it with a bounded worker, and verify it with evaluator and error-checker roles."

plugins/subagent-orchestration-kit/contexts/hackhub-product-overview.md

@@ -0,0 +1,85 @@
+# HackHub Product Overview
+
+## Source of truth order
+
+Use this order when the repository is inconsistent:
+
+1. Direct user instructions in the current thread
+2. This file
+3. The current repo product and architecture docs
+4. Any sibling-repo specs that this product explicitly inherits from
+5. Current implementation
+
+## Current workspace path mapping
+
+This context file is written for the current HackHub workspace layout.
+
+From the `changeHack/` repo root, the intended doc locations are:
+
+- `docs/`
+- `../progress-checker/docs/`
+- `../hack-evaluater/docs/`
+
+If those sibling paths do not exist in another repo, do not follow them. Replace this file with a repo-specific overview derived from [project-overview-template.md](./project-overview-template.md) first.
+
+## Product direction
+
+The intended product is not just the current `changeHack` MVP.
+
+It is a combined product that should merge:
+
+- the Slack-driven progress and question workflow from `progress-checker`
+- the multi-hackathon and role-aware product model from `hack-evaluater`
+
+The product must support multiple hackathons and serve:
+
+- participants / hackers
+- mentors / judges
+- organizers
+- corporate sponsors
+- tenant or platform admins when needed
+
+## Problem the product solves
+
+- Participants need a low-friction way to ask questions and share progress.
+- Mentors need structured, triaged questions instead of unformatted Slack messages.
+- Organizers need a web view of progress, blockers, and unanswered questions across hackathons.
+- Sponsors need scoped access to the hackathons and teams they are allowed to see.
+
+## Combined product shape
+
+### From progress-checker
+
+- `/progress` modal and progress board flow
+- `/question` modal and AI first response flow
+- async queue architecture for Slack's 3-second constraint
+- web dashboard for progress and questions
+- future Slack discussion to GitHub issue flow
+
+### From hack-evaluater
+
+- multi-hackathon structure
+- role separation and scoped permissions
+- web architecture patterns for feature/container based UI
+- backend layering and clearer domain boundaries
+- sponsor and organizer facing workflows
+
+## Known current drift
+
+The current `changeHack` implementation is still closer to a KCL-specific support hub MVP:
+
+- single-event assumptions remain in docs and code
+- sponsor and organizer flows are not implemented
+- multi-hackathon modeling is not implemented
+- mentor-only operational views are incomplete
+- Slack discussion to Issue flow is still missing
+
+Treat these as implementation gaps, not final product decisions.
+
+## Planning constraints
+
+- Preserve Slack-first entry points.
+- Keep heavy work asynchronous.
+- Do not break the existing `changeHack` MVP while expanding scope.
+- Favor small vertical slices over broad rewrites.
+- When docs conflict, note the conflict explicitly before choosing a path.

plugins/subagent-orchestration-kit/contexts/project-overview-template.md

@@ -0,0 +1,40 @@
+# Project Overview Template
+
+## Source of truth order
+
+1. Current user instructions
+2. This file
+3. Primary product specs
+4. Architecture and permission docs
+5. Current implementation
+
+## Product direction
+
+- What the product is
+- Which older repos or docs it merges
+- Which user roles matter most
+- Which scope assumptions are critical
+
+## Target users
+
+- user group 1
+- user group 2
+- user group 3
+
+## Core workflows
+
+- workflow 1
+- workflow 2
+- workflow 3
+
+## Non-negotiable constraints
+
+- architecture constraint
+- security constraint
+- product constraint
+- delivery constraint
+
+## Known drift
+
+- docs vs implementation mismatch 1
+- docs vs implementation mismatch 2

plugins/subagent-orchestration-kit/skills/coding-worker/SKILL.md

@@ -0,0 +1,39 @@
+---
+name: coding-worker
+description: Implement one bounded slice with clear ownership, tests, and minimal scope creep.
+---
+
+# Coding Worker
+
+## Overview
+
+Use this skill for implementation once the orchestrator has already chosen a slice.
+
+This role owns one bounded write scope and should avoid expanding the task.
+
+## Required inputs
+
+- exact acceptance criteria
+- the files or modules this role owns
+- known constraints from the project overview
+  - in this repo: `../../contexts/hackhub-product-overview.md`
+- any test requirements for the slice
+
+## Required behavior
+
+- Implement only the assigned slice.
+- Preserve existing behavior outside the slice.
+- Add or update tests for the touched area.
+- Surface blockers instead of silently changing scope.
+
+## Deliverables
+
+- code changes
+- tests or verification updates
+- short note on any unresolved risk
+
+## Guardrails
+
+- Do not redefine the product while implementing.
+- Do not take ownership of files assigned to another worker.
+- Prefer small diffs over broad refactors unless the slice explicitly requires one.

plugins/subagent-orchestration-kit/skills/error-checker/SKILL.md

@@ -0,0 +1,69 @@
+---
+name: error-checker
+description: Compare the repository's code and plans against the actual product direction, specs, and architectural constraints.
+---
+
+# Error Checker
+
+## Overview
+
+Use this skill to find implementation drift.
+
+This role is not only a bug checker. It checks whether the repository is moving in the wrong direction.
+
+## Required inputs
+
+- the project overview file
+  - in this repo: `../../contexts/hackhub-product-overview.md`
+  - in another repo: the project-specific overview that replaced it
+- relevant specs and architecture docs
+- the code or plan under review
+
+## What to look for
+
+### Product drift
+
+- current implementation assumes a smaller product than the docs require
+- single-event assumptions where multi-hackathon support is required
+- mentor-only assumptions where sponsor or organizer workflows are required
+
+### Architectural drift
+
+- synchronous handling where async design is required
+- permission shortcuts that violate the intended role model
+- hard-coded assumptions that block future role or tenant expansion
+
+### Implementation gaps
+
+- documented flows with no code path
+- partial flows with no persistence or no UI
+- UI routes that exist without backend support
+- backend support that exists without user-facing flow
+
+### Quality problems
+
+- missing tests in modified areas
+- broken or stale docs after implementation
+- hidden coupling and poor ownership boundaries
+
+## Output format
+
+List findings in severity order:
+
+1. `critical`
+2. `high`
+3. `medium`
+4. `low`
+
+Each finding should include:
+
+- what is wrong
+- why it matters
+- where it shows up
+- the narrowest fix that would reduce the problem
+
+## Guardrails
+
+- Do not treat every future feature as a bug.
+- Focus on real mismatches between product direction, docs, and implementation.
+- Flag contradictions explicitly when the source docs disagree.