The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the Hex Graph listing page.
A compact marketplace of standalone engineering skills for Claude Code and Codex.
Supports the portable Agent Plugins v1 standard while retaining native Claude Code and Codex distribution compatibility.
Why this repository is intentionally small: Each skill provides domain knowledge, decision gates, tool guidance, and evidence checks for one bounded outcome. Instructions that duplicate an existing obligation or do not support that outcome are removed; useful checks remain individually accountable.
Browse the skills catalog or install only the plugins you need below.
The repository intentionally contains only the skills, minimal plugin manifests, two host-specific marketplace catalogs, documentation, and a static catalog site. It has no MCP servers, orchestration hierarchy, distributed shared resources, generated skill copies, or evaluation harness.
The repository skill template defines the shared authoring format. Skills track every domain checklist item with evidence or an explicit applicability reason. A final self-check reconciles evidence, scope, omissions, and verdict. Reports use the same order—result, scope, evidence, verification, completion—with skill-specific decisions and evidence. Supporting references stay inside the skill and are loaded only for the relevant workflow branch; the delivery review panel is one such branch. Skills have a 200-line ceiling and no minimum.
| Index | Skill | Purpose |
|---|---|---|
| 11 | Plan Reviewer | Validate an implementation plan against repository evidence before execution. |
| 12 | Delivery Reviewer | Review scoped delivery acceptance, regressions, and release risks; use independent perspectives only when needed and permitted. |
| Index | Skill | Purpose |
|---|---|---|
| 21 | Documentation Auditor | Audit documentation trust, coverage, structure, and factual accuracy. |
| 22 | Codebase Auditor | Audit cross-cutting code health, security, delivery, and maintainability. |
| 23 | Test Suite Auditor | Audit test value, coverage, isolation, and oracle strength. |
| 24 | Architecture Auditor | Audit system boundaries, ownership, contracts, and dependency topology. |
| 25 | Persistence Auditor | Audit queries, transactions, consistency, and resource lifecycle. |
| Index | Skill | Purpose |
|---|---|---|
| 31 | Performance Optimizer | Profile, experiment, and keep only measured improvements. |
| 32 | Dependency Upgrader | Upgrade dependencies in reversible, verified batches. |
| 33 | Code Modernizer | Replace or simplify bounded capabilities when net value is proven. |
| 34 | Benchmark Comparator | Compare alternatives with a frozen, reproducible experiment contract. |
| 35 | Surgical Change Implementer | Implement a bounded change through the smallest complete root-cause solution. |
| Index | Skill | Purpose |
|---|---|---|
| 41 | Test Strategy Planner | Design a risk-based, decision-complete test strategy without changing code. |
| 42 | Acceptance Test Builder | Build, update, consolidate, or retire acceptance evidence without repairing product code. |
| Index | Skill | Purpose |
|---|---|---|
| 51 | Opportunity Evaluator | Compare product opportunities using current evidence and a low-cost validation path. |
Optional maintainer toolkit: skill 61 reviews skill repositories, skill 62 publishes any Git repository with equivalent remote evidence, and skills 63–64 publish GitHub releases and Discussions. Users who only consume the engineering skills do not need to install it.
| Index | Skill | Purpose |
|---|---|---|
| 61 | Skill Reviewer | Review skill content, goal alignment, triggers, and distribution contracts; distinguish static review from authorized behavioral validation. |
| 62 | Repository Publisher | Validate, commit, push, and remotely verify approved repository changes. |
| 63 | Release Publisher | Prepare and publish an approved tagged GitHub release. |
| 64 | Community Announcer | Draft and publish fact-checked GitHub Discussions project announcements. |
Creates durable architecture artifacts without coupling the skills to one another. Each workflow operates independently and communicates only through optional repository documents.
| Index | Skill | Purpose |
|---|---|---|
| 71 | System Design Baseline Builder | Establish measurable architecture drivers and constraints in one project baseline. |
| 72 | Current Architecture Documenter | Document implemented architecture from repository evidence. |
| 73 | System Design Proposal Builder | Turn requirements and constraints into a decision-complete target design. |
| 74 | Architecture Decision Recorder | Record one significant decision with alternatives and consequences. |
| 75 | Architecture Diagram Builder | Create evidence-backed current or target architecture views. |
| 76 | Architecture Migration Planner | Plan a safe current-to-target transition with compatibility, reversibility limits, and recovery actions. |
Add the marketplace and install only the suites you need:
Invoke a skill by its namespaced name, for example /review-suite:ln-12-delivery-reviewer.
For local development, load one plugin directly:
Each plugin contains a portable Agent Plugins v1 plugin.json, the current .codex-plugin/plugin.json OpenAI host adapter, and one shared skills/<skill>/SKILL.md tree.
This is the smallest practical shared layout for distributed plugins:
skills/<name>/SKILL.md, so each skill has one canonical copy.plugin.json; its minimal manifest owns only the schema target and stable name..codex-plugin/plugin.json, which remains the single owner of mutable version, description, publisher, and interface metadata.skills/ directory, so a duplicate Claude-specific plugin manifest is unnecessary.SKILL.md. Scripts or assets are added only for a concrete execution or output need; no shared runtime or MCP dependency is required.The structure follows the Agent Plugins v1 specification, current OpenAI plugin guide, Agent Skills specification, Claude Code skill guide, and Claude Code plugin reference.
The first digit identifies the plugin and the second identifies the skill within it: 1x review, 2x audit, 3x optimization, 4x testing, 5x product discovery, 6x repository maintenance, and 7x architecture artifact creation. See the canonical allocation and overflow rules in AGENTS.md.
Use SKILL_TEMPLATE.md for authoring and AGENTS.md for repository rules and required validation. Review each instruction against the skill's goal: remove unnecessary work, fill missing evidence or safety obligations, and clarify conditions, ordering, and failure handling. Preserve domain checks rather than replacing an entire phase with one checkbox.
Run pwsh -File scripts/validate-repository.ps1 for repository structure, common skill contracts, catalogs, and documentation coverage, together with the additional validators required by AGENTS.md. Static validation establishes format and consistency; it does not certify task performance. Ordinary edits do not change plugin versions or create a release.