merceralex397-collab

Find external skills from public registries, GitHub repos, and official skill collections, then evaluate them for quality, licensing, and fitness for adoption. Use when looking for existing skills before building from scratch, evaluating external skill quality, or migrating community skills into a local library. Do not use when building a novel skill with no external precedent or for quick one-off evaluation (just read the skill directly).

Improve an existing skill package — tighten routing, sharpen procedure, add or prune support layers, upgrade packaging. Use when the user says "improve this skill", "this skill is weak/vague/bloated", "harden this SKILL.md", or "add evals/references to this skill package". Do not use for creating a new skill from scratch (use skill-creator), trigger-only fixes when the body is fine (use skill-trigger-optimization), porting a skill to a different stack or context (use skill-adaptation), or quick structural audits with no rewrite (use skill-anti-patterns).

Install a skill package into the local agent client from a GitHub repository, local folder, or archive. Use when the user says "install this skill", "add skill from GitHub", or "list available skills". Do not use for creating new skills (use skill-creator), packaging skills for distribution (use skill-packaging), or improving already-installed skills (use skill-improver).

Orchestrate multi-skill pipelines through the CLI. Use when "run the creation pipeline", "execute the improvement workflow", "resume my pipeline", or when chaining multiple skills with decision points. Supports Creation, Improvement, and Library Management pipelines with state persistence and conditional branching. Do not use for single skill operations (use the skill directly) or exploratory tasks without defined workflow.

Bundle one or more completed skill folders into versioned distributable archives with manifests, integrity checksums, and OpenCode metadata. Use when a user says "package this skill", "bundle for distribution", "prepare a versioned release", "generate OpenCode metadata", "build a release bundle", or "package these skills for release". Do not use for installing bundles (use skill-installer), writing new skills (use skill-creator), or documenting skill origin and trust chain (use skill-provenance).

Polish and finalize an AI agent prompt file by refining structure, wording, and clarity to match proven best practices while preserving the original intent and markdown frontmatter.

Integrate LLM capabilities into applications with explicit runtime boundaries, structured schemas, cost controls, and evaluation plans. Triggers on tasks involving model inference APIs, prompt engineering systems, LLM toolchain setup, or AI agent runtime design. Does not trigger on ordinary software tasks without model, inference, evaluation, or agent-runtime concerns.

Register, version, and govern MLflow models through a full lifecycle from training-run artifact to production stage with validation gates, alias-based deployment routing, lineage tracking, and CI/CD automation. Triggers on "register a model", "promote model to production", "manage model versions", "model stage transitions", "model governance", "model rollback", "model lineage", and "champion/challenger A/B models".

Design and harden agent, command, workflow, and tool prompts for reliable execution across different AI models. Use when creating or revising repo-local agents to apply model-specific prompting techniques, tighten scope, and prevent common agent failure modes like doom loops, status-over-evidence routing, and impossible read-only delegation. Do not use for one-off prompts or when agents are already working reliably.

Find external skills from public registries, GitHub repos, and official skill collections, then evaluate them for quality, licensing, and fitness for adoption. Use when looking for existing skills before building from scratch, evaluating external skill quality, or migrating community skills into a local library. Do not use when building a novel skill with no external precedent or for quick one-off evaluation (just read the skill directly).

Create concise project handoff artifacts, especially a top-level START-HERE document and short resume context for the next session or machine. Use when scaffolding finishes, a milestone closes, or long-running autonomous work needs a compact restart surface. Do not use when work is fully complete with nothing pending, or for loading context at session start.

Generate structured migration plans with inventory, risk-ranked sequence, rollback procedures, and acceptance criteria for major upgrades or framework migrations. Use when upgrading major dependencies (React 17→18, Python 3.9→3.12), migrating frameworks (Express→Fastify, CRA→Vite), or scoping 'modernize the codebase' initiatives. Do not use for patch-level dependency updates or greenfield projects.

Design and generate a project-specific agent team with specialized agents, tools, plugins, commands, and skills tailored to the project type and stack. Use after the base scaffold exists to customize generic agent templates into project-aware specialists. Do not use before a scaffold exists (run scaffold-kickoff first) or when the project is too simple for agent specialization.

Generate client-specific overlay files (metadata, permissions, manifests) from a canonical SKILL.md source for different AI agent clients. Use when publishing a skill to multiple ecosystems (Copilot, OpenCode, Codex, Gemini CLI), converting between overlay formats, or validating overlay consistency. Do not use when creating skills for a single client only or when editing skill content (edit SKILL.md first, then regenerate overlays).

Review a pull request, validate review comments against the actual implementation, and generate follow-up tickets only for findings that survive evidence-based triage. Use after a PR review cycle when comments need to be triaged into actionable work items. Do not use during an in-progress review, when all comments were already addressed in the PR, or for discussions that aren't action items.

Create project-local skills populated with actual project data, stack-specific conventions, and domain-specific procedures. Use after scaffolding to replace generic skill placeholders with real project-aware guidance that helps agents work effectively in this specific repo. Do not use before the project has established patterns (scaffold first) or for universal skills that belong in a skill library.

Design and configure the asset sourcing, generation, provenance, compliance, optimization, and import pipeline for game projects. Use after repo-scaffold-factory when a generated repo needs machine-checkable asset truth.

Hold the project-local standards for languages, frameworks, validation, and runtime assumptions. Use when planning or implementing work that should follow repo-specific engineering conventions.

Follow the required ticket lifecycle for this repo. Use when an agent is advancing a ticket through planning, review, implementation, QA, deterministic smoke testing, and closeout and needs the repo-specific stage rules.

Generate the base repository file structure including README, AGENTS.md, docs layout, ticket templates, and the full OpenCode scaffold with agents, tools, plugins, commands, and skills. Use when creating a greenfield repo foundation. This generates a generic starting structure that other skills then customize.

Run Scafforge's host-side diagnosis flow for an existing repository. Use when a repo needs workflow diagnosis, contract verification, professional codebase review, or a four-report diagnosis pack with evidence-backed ticket recommendations and no repair edits.

Keep code review, security review, and QA passes evidence-based and stage-aware for this repo. Use when a ticket already has an approved plan and implementation evidence and a review lane needs structured findings.

Apply Scafforge's host-side managed repair flow for an existing repository. Use when diagnosis has already identified safe workflow repairs, managed-surface replacement, or contract-refresh work that should be applied with provenance and post-repair verification.

Design and harden agent, command, workflow, and tool prompts for reliable execution across different AI models. Use when creating or revising repo-local agents to apply model-specific prompting techniques, tighten scope, and prevent common agent failure modes like doom loops, status-over-evidence routing, and impossible read-only delegation.

Run browser-game playtests and frontend QA. Use when the user asks for smoke tests, screenshot-based verification, browser automation, HUD or overlay review, or structured issue-finding in a browser game.

Route early browser-game work. Use when the user needs stack selection and workflow planning across design, implementation, assets, and playtesting before moving to a specialist skill.

Design UI surfaces for browser games. Use when the user asks for HUDs, menus, overlays, responsive layouts, or visual direction that must protect the playfield.

Implement 2D browser games with Phaser. Use when the user wants a Phaser, TypeScript, and Vite stack for scenes, gameplay systems, cameras, sprite animation, and DOM-overlay HUD patterns.

Build React-hosted 3D browser games with React Three Fiber. Use when the user wants pmndrs-based scene composition, shared React state, and 3D HUD integration inside a React app.

Generate and normalize 2D sprite animations. Use when the user asks for full-strip generation from approved source frames, consistent anchor and scale normalization, or preview assets for browser-game animation.

Implement browser-game runtimes with plain Three.js. Use when the user wants imperative scene control in TypeScript or Vite with GLB assets, loaders, physics, and low-level WebGL debugging.

Prepare and optimize browser-game 3D assets. Use when the user asks for GLB or glTF shipping work, including Blender cleanup and export, collision or LOD setup, compression, texture packaging, and runtime validation.

Hold the project-local standards for languages, frameworks, validation, and runtime assumptions. Use when planning or implementing work that should follow repo-specific engineering conventions.

Follow the required ticket lifecycle for this repo. Use when an agent is advancing a ticket through planning, review, implementation, QA, deterministic smoke testing, and closeout and needs the repo-specific stage rules.

Apply the `Weak-tier evidence-first profile` operating profile for the selected downstream models. Use when shaping prompts, delegation briefs, review asks, or evidence requests for this repo.

Keep code review, security review, and QA passes evidence-based and stage-aware for this repo. Use when a ticket already has an approved plan and implementation evidence and a review lane needs structured findings.

Hold GPTTalker’s Qdrant, indexing, issue-memory, and cross-repo intelligence conventions.

Hold GPTTalker-specific MCP hub rules for tool contracts, policy enforcement, and fail-closed behavior.

Hold GPTTalker’s node-agent design rules for local execution, Tailscale transport, and bounded subprocess work.

Re-run repo-process-doctor safely after scaffold or workflow changes and record what was replaced versus preserved.

Repo-specific Godot 4.6 Android standards for Woman vs Horse VB, including free/open asset provenance and credits requirements.

Hold the project-local standards for languages, frameworks, validation, and runtime assumptions. Use when planning or implementing work that should follow repo-specific engineering conventions.

Follow the required ticket lifecycle for this repo. Use when an agent is advancing a ticket through planning, review, implementation, QA, deterministic smoke testing, and closeout and needs the repo-specific stage rules.

Hold the repo-specific Godot, Android-export, asset-provenance, and validation rules for Woman vs Horse VB.

Follow the required ticket lifecycle for this repo. Use when an agent is advancing a ticket through planning, review, implementation, QA, deterministic smoke testing, and closeout and needs the repo-specific stage rules.

How to find, evaluate, download, and track CC0/CC-BY game assets from Kenney.nl, OpenGameArt.org, and Freesound.org. Use when sourcing sprites, tilesets, audio, or fonts for this project.

Godot 4.6 Android game development patterns for GDScript, scene structure, asset import, touch input, and export. Use when planning or implementing Godot scenes, scripts, or Android builds.

Keep code review, security review, and QA passes evidence-based and stage-aware for this repo. Use when a ticket already has an approved plan and implementation evidence and a review lane needs structured findings.

Hold the project-local standards for languages, frameworks, validation, and runtime assumptions. Use when planning or implementing work that should follow repo-specific engineering conventions.

Follow the required ticket lifecycle for this repo. Use when an agent is advancing a ticket through planning, review, implementation, QA, deterministic smoke testing, and closeout and needs the repo-specific stage rules.

Apply the `Standard-tier evidence-first profile` operating profile for the selected downstream models. Use when shaping prompts, delegation briefs, review asks, or evidence requests for this repo.

Godot 4.6 Android game development patterns, scene architecture, GDScript conventions, touch input, and export configuration. Use when planning or implementing any game scene, script, or Android build work.

GDScript procedural sprite generation patterns for Route A (colored shapes, no external assets). Use when implementing any visual element — player, enemies, UI, particles, or arena.

Guide an agent through writing a precise, actionable asset description brief that can be executed by a Blender-MCP subagent or used to source matching free assets. Use when a game project needs a specific visual asset defined before creation.

Guide agents through the repo's managed Blender-MCP asset workflow, including saved-blend chaining and audit-log verification.

Hold Godot 4.6 3D game patterns for this Android arena game. Use when implementing scenes, importing GLB models, setting up orthographic camera, or writing GDScript for 3D gameplay.

Hold the repo-specific engineering, validation, and release-proof standards for Woman vs Horse VC's Godot 4.6 Android stack.

Follow the required ticket lifecycle for this repo. Use when an agent is advancing a ticket through planning, review, implementation, QA, deterministic smoke testing, and closeout and needs the repo-specific stage rules.

Repo-specific Godot 4.6 Android standards for Woman vs Horse VD, including zero-import Godot-native art and procedural SFX rules.

Godot 4.6 Android game development patterns, project structure, and mobile-specific guidance. Use when planning or implementing Godot scenes, scripts, exports, or touch input.

How to create all game visuals using only Godot built-in features. No external asset imports allowed. Covers Polygon2D, ShaderMaterial, GPUParticles2D, TileMap, Theme, Line2D, and AudioStreamGenerator with GDScript examples for each technique.

View and modify Codex plugin settings for this session

Unified Codex command with auto-detection, intent routing, and quality gates