research-foundry

Drives the Research Foundry control plane — a Markdown/YAML-first research pipeline that captures an idea, routes it through source discovery, claim mapping, synthesis, and deterministic verification, then publishes an evidence bundle. Use when the operator wants to run, plan, or audit an end-to-end research run with the `rf` CLI, or wants claim-traceable, governance-gated research instead of a freehand answer.

Installs and initializes a Research Foundry workspace from scratch, then orchestrates a Claude Code agent swarm to produce a claim-verified evidence bundle. Covers the uv-tool install, `rf init`, `rf doctor` bootstrap sequence; the two swarm-drive patterns (RF-native `rf swarm run` vs. Claude-Code-orchestrated discovery feeding `rf ingest`); the deterministic tail that converts raw sources into a verified, governance-gated report; and key-profile governance. Cross-links the `research-foundry` skill for the full 21-step pipeline loop and per-command reference. Use when an agent needs to stand up a Research Foundry workspace from scratch or orchestrate a Claude Code agent swarm into a claim-verified evidence bundle.

Unified execution engine for all development workflows. Progressive disclosure for phase execution, quick features, story completion, and scaffolding. Integrates with artifact-tracking and meatycapture-capture. Use when running /dev:execute-phase, /dev:quick-feature, /dev:implement-story, /dev:complete-user-story, or /dev:create-feature commands.

Generate and optimize PRDs, Implementation Plans, and Progress Tracking documents optimized as AI artifacts for development agents. Use when creating new feature plans, breaking down long planning docs (>800 lines), or setting up progress tracking. Supports: 1) Create PRD from feature request, 2) Create Implementation Plan from PRD with phase breakdown and subagent assignments, 3) Optimize existing plans by breaking into phase-specific files, 4) Create progress tracking with task assignments. Example: "Create a PRD for user authentication feature" or "Break down the sidebar-polish implementation plan into phase files" or "Create progress tracking for data-layer-fixes PRD".

Token-efficient codebase navigation through intelligent symbol loading and querying. Use this skill when implementing new features (find existing patterns), exploring codebase structure, searching for components/functions/types, or understanding architectural layers. Reduces token usage by 60-95% compared to loading full files. Layer split enables 50-80% additional backend efficiency.

Token-efficient tracking for AI orchestration. CLI-first for status updates (~50 tokens), agent fallback for complex ops (~1KB). Use when: updating task status, querying blockers, creating progress files, validating phases.

TypeScript authentication framework (framework-agnostic). Features: email/password, OAuth (Google, GitHub, Discord), 2FA (TOTP, SMS), passkeys/WebAuthn, session management, RBAC, rate limiting, database adapters. Actions: implement, configure, secure authentication systems. Keywords: Better Auth, authentication, authorization, OAuth, email/password, 2FA, MFA, TOTP, passkeys, WebAuthn, session management, RBAC, rate limiting, database adapter, TypeScript auth, social login, Google auth, GitHub auth, Discord auth, email verification, password reset. Use when: implementing TypeScript auth, adding OAuth providers, setting up 2FA/MFA, managing sessions, configuring RBAC, building secure auth systems.

Delegate bounded subtasks to IBM Bob Shell CLI. Use when the user wants to use Bob, delegate to Bob, run Bob for drafting/scaffolding/exploration, or compare Codex work with Bob output. Do NOT use for architecture-heavy integration work or cross-layer backend changes.

Browser automation, debugging, and performance analysis using Puppeteer CLI scripts. Use for automating browsers, taking screenshots, analyzing performance, monitoring network traffic, web scraping, form automation, and JavaScript debugging.

Expert reference for Codex itself — features, setup, troubleshooting, and enterprise use. Use this skill when answering questions about Codex installation, authentication, slash commands, agent skills, MCP server configuration, hooks, plugins, IDE integration (VS Code, JetBrains), CI/CD integration, enterprise deployment (SSO, sandboxing, monitoring), extended thinking, caching, checkpointing, cost tracking, or debugging Codex behaviour. Distinct from the user-level Extensibility skill: this skill is about operating and understanding Codex, not about authoring new extensions.

Use when the user asks to run Codex CLI (codex exec, codex resume) or references OpenAI Codex for code analysis, refactoring, or automated editing. Uses GPT-5.3-Codex by default for state-of-the-art software engineering.

Use when designing visual interfaces, data visualizations, educational content, or presentations and need to ensure they align with how humans naturally perceive, process, and remember information. Invoke when user mentions cognitive load, visual hierarchy, dashboard design, form design, e-learning, infographics, or wants to improve clarity and reduce user confusion. Also applies when evaluating existing designs for cognitive alignment or choosing between design alternatives.

Pre-implementation confidence assessment (≥90% required). Use before starting any implementation to verify readiness with duplicate check, architecture compliance, official docs verification, OSS references, and root cause identification.

Run an Agent Review Council over a product, workflow, codebase, architecture, prompt, agent package, or GTM artifact. Produces evidence-backed structured findings, scorecard, risks, decision record, and validation plan.

Scaffold an ARC run, choose or recommend a council, and author or edit council definitions and reviewer roles. Triggers on starting/scaffolding a run, picking the right council for an objective, creating/editing council YAML or reviewer-role YAML, rendering agent stubs, using the optional agentic compose surface, and setting up project-specific councils with arc project init, arc councils resolve, --project/--council-type flags, or SAM modes — via the arc CLI, the Web portal, or the HTTP API. Use council-review (not this skill) to actually run the review.

Debug and remediate bugs with severity-gated workflows. Progressive disclosure from quick triage through comprehensive remediation planning. Integrates with planning skill for complex fixes and artifact-tracking for progress. Also provides a post-incident retrospective mode: classifies post-merge gaps against a failure taxonomy and emits a patch-plan addendum. Use when running /fix:debug or /fix:bugfix-commit commands. /fix:debug auto-invokes the retrospective workflow when the prompt contains: 'Codex had to patch', 'gaps after merge', 'we missed', or 'regression after phase'.

Unified execution engine for all development workflows. Progressive disclosure for phase execution, quick features, story completion, and scaffolding. Integrates with artifact-tracking and meatycapture-capture. Use when running /dev:execute-phase, /dev:quick-feature, /dev:implement-story, /dev:complete-user-story, or /dev:create-feature commands.

Create distinctive, production-grade frontend interfaces with high design quality. Use this skill when the user asks to build web components, pages, or applications. Generates creative, polished code that avoids generic AI aesthetics.

Google Gemini CLI orchestration for AI-assisted development. Capabilities: second opinion/cross-validation, real-time web search (Google Search), codebase architecture analysis, parallel code generation, code review from different perspective, image generation, SVG/animation generation, multimodal screenshot-to-code. Actions: query, search, analyze, generate, review, visualize with Gemini. Keywords: Gemini CLI, second opinion, cross-validation, Google Search, web research, current information, parallel AI, code review, architecture analysis, gemini prompt, AI comparison, real-time search, alternative perspective, code generation, image gen, UI mockup, SVG, animation, multimodal. Use when: needing second AI opinion, searching current web information, analyzing codebase architecture, generating code in parallel, getting alternative code review, researching current events/docs, generating UI mockups, creating SVG animations, screenshot-to-code workflows.

Drive the IntentTree work-tree through the `intenttree`/`itt` CLI via natural language. Use to capture inbox items, create or decompose nodes (pillar → work_area → work_package → atomic_task → step), query "what's next", complete/update/assign/move/defer nodes, run the M1 copy/paste agent dispatch loop, pull prior context, and bootstrap IntentTree into a project. Triggers: "log this to IntentTree", "create a task/tree", "what should I work on next", "schedule today", "start a run", "decompose this", "mark it done". Requires a running API server; the CLI is a thin httpx REST client, never a direct DB accessor. Progressive disclosure: load only the one workflow doc the intent matches.

Create, rebuild, or maintain a project README using a modular Handlebars build system. Covers bootstrapping the .github/readme/ scaffold for any project, planning screenshot and GIF visual assets, crafting content strategy (audience analysis, section structure, success metrics), assembling or rebuilding README.md from partials and data files, validating links and screenshots, and wiring CI workflows or pre-commit hooks for README freshness. Also use when a version bump, feature addition, or screenshot update requires regenerating README content. Supports CLI tools, web apps, libraries, and SaaS products.

Use this skill when authoring markdown artifacts destined for MeatyWiki ingestion. Guides agents in writing valid YAML frontmatter with routing hints that the engine honors during ingest → classify → compile. Triggers: "write a MeatyWiki artifact", "author an artifact for the wiki", "create a concept/research-finding/code-snippet for MeatyWiki", "add frontmatter hints", "meatywiki frontmatter", "routing hints", "meatywiki-author". Do NOT use for: authoring general markdown, running meatywiki CLI commands (use meatywiki skill), Portal UI actions.

Drive the MeatyWiki CLI end-to-end — knowledge-compilation loop (ingest → classify → extract → compile → file-back → lint) over a file-first Obsidian-compatible markdown vault. Use when running any MeatyWiki CLI task, compiling inbox sources, querying/searching the wiki (including semantic/vector search), checking lens scores, managing workflow runs, evaluating decisions, OR bootstrapping MeatyWiki into a fresh project (vault init + AGENTS.md sidecar + wiki-spec stub). Triggers: "meatywiki", "compile loop", "ingest into wiki", "vault", "knowledge compilation", "semantic search", "lens score", "fidelity", "freshness", "workflow run", "decision table", "bootstrap meatywiki", "set up meatywiki", "wire meatywiki into this project". Do NOT use for: non-MeatyWiki knowledge systems, Portal UI, SAM/CCDash (those are separate skills/projects).

Full-lifecycle MeatyWiki usage — compilation engine CLI, Portal REST API, and MCP server. Use when: running any MeatyWiki CLI command, calling Portal API endpoints, performing vault CRUD, ingesting content, compiling artifacts, querying knowledge, managing workflows, or deploying MeatyWiki for agentic use.

Generate/edit images with Nano Banana Pro (Gemini 3 Pro Image). Use for image create/modify requests incl. edits. Supports text-to-image + image-to-image; 1K/2K/4K; use --input-image.

Generates AI images using the nano-banana CLI (Gemini 3.1 Flash default, Pro available). Handles multi-resolution (512-4K), aspect ratios, reference images for style transfer, green screen workflow for transparent assets, cost tracking, and exact dimension control. Use when asked to "generate an image", "create a sprite", "make an asset", "generate artwork", or any image generation task for UI mockups, game assets, videos, or marketing materials.

Complete API for Google NotebookLM - full programmatic access including features not in the web UI. Create notebooks, add sources, generate all artifact types, download in multiple formats. Activates on explicit /notebooklm or intent like "create a podcast about X"

Use this skill to query your Google NotebookLM notebooks directly from Codex for source-grounded, citation-backed answers from Gemini. Browser automation, library management, persistent auth. Drastically reduced hallucinations through document-only responses.

Deploy and manage NotebookLM documentation sync for any Codex project. Installs Python scripts + Codex hook that automatically syncs markdown files to a Google NotebookLM notebook on every Write/Edit. Supports install, status, resync, batch sync, cleanup, and uninstall workflows. Use when setting up NotebookLM sync in a new project, checking sync health, or managing notebook sources.

Post-implementation retrospective on a completed implementation plan. Compares estimated story points against measured complexity (git stats, file counts, CCDash AAR), identifies which estimation heuristics failed, and emits actionable tuning suggestions for the planning skill. Use when a feature is done or near-done and the user asks for a retrospective ("why did X take so long", "review the plan for X", "AAR on X", "what did we miss in the estimate"), or proactively after any plan ≥5 pts completes. Complements CCDash's `feature-retrospective` recipe by adding heuristic-tuning analysis specific to SkillMeat's planning skill.

Cross-feature visibility layer for implementation plans, PRDs, and progress files. Use when: checking what's in progress, auditing stale plan statuses, remediating mismatches between frontmatter and task reality, summarizing active work, investigating orphaned/superseded PRDs. Complements artifact-tracking (which handles per-phase task updates). Routes intent to scripts/plan-status-report.py and artifact-tracking scripts.

Generate and optimize PRDs, Implementation Plans, and Progress Tracking documents optimized as AI artifacts for development agents. Use when creating new feature plans, breaking down long planning docs (>800 lines), or setting up progress tracking. Supports: 1) Create PRD from feature request, 2) Create Implementation Plan from PRD with phase breakdown and subagent assignments, 3) Optimize existing plans by breaking into phase-specific files, 4) Create progress tracking with task assignments. Example: "Create a PRD for user authentication feature" or "Break down the sidebar-polish implementation plan into phase files" or "Create progress tracking for data-layer-fixes PRD".

Produce changelog-first, evidence-backed summaries of prior-day work for prompts like "what we did yesterday", "what shipped yesterday", "what changed yesterday", "what was visibly different", and "how do I try it". Use this skill for daily automation reports, release-adjacent status synthesis, empty-day audits, and branch-local versus shipped reconciliation. Progressive disclosure: classify the day first, then load only the workflow docs needed for visible-surface analysis, debt/wins framing, or visual-summary output.

Use this skill to distill a software project (codebase plus optional supplemental docs, ADRs, PRDs, NotebookLM exports, design/business notes) into high-signal, research-ready context artifacts. Produces four synthesized markdown documents — feature catalog, design fundamentals, research context pack, and opportunity map — grounded in repo evidence with explicit confidence levels, intended for downstream AI agents reasoning about expansion, novel ideas, or architectural evolution. Trigger when the user asks to "distill project context", "generate project understanding artifacts", "externalize project context for future agents", "build a research context pack", or points the skill at a repo with supplemental materials.

Use this skill when scaffolding a new project or feature from context (PRD, implementation plan, project directory, or free-text description). Transforms context into a deployed Bundle of artifacts via structured Intent extraction and ranked artifact search. Invoked by /scaffold:project and /scaffold:feature commands. Trigger phrases: 'scaffold my project', 'set up my Codex environment', 'create a starter bundle', 'scaffold from my PRD', 'build my artifact bundle', 'feature scaffold', 'scaffold feature'. Requires a seeded local collection (≥20 artifacts) and a running SkillMeat API.

Recover from crashed, failed, or interrupted Codex sessions. Use this skill when: session crashed during multi-agent parallel execution, need to determine what work was completed vs incomplete, want to generate resumption commands for interrupted tasks, recovering from context window exhaustion, or handling session handoffs. Analyzes agent logs, verifies on-disk state, and creates resumption plans with ready-to-execute Task() commands.

Drives the Research Foundry control plane — a Markdown/YAML-first research pipeline that captures an idea, routes it through source discovery, claim mapping, synthesis, and deterministic verification, then publishes an evidence bundle. Use when the operator wants to run, plan, or audit an end-to-end research run with the `rf` CLI, or wants claim-traceable, governance-gated research instead of a freehand answer.

Installs and initializes a Research Foundry workspace from scratch, then orchestrates a Codex agent swarm to produce a claim-verified evidence bundle. Covers the uv-tool install, `rf init`, `rf doctor` bootstrap sequence; the two swarm-drive patterns (RF-native `rf swarm run` vs. Codex-orchestrated discovery feeding `rf ingest`); the deterministic tail that converts raw sources into a verified, governance-gated report; and key-profile governance. Cross-links the `research-foundry` skill for the full 21-step pipeline loop and per-command reference. Use when an agent needs to stand up a Research Foundry workspace from scratch or orchestrate a Codex agent swarm into a claim-verified evidence bundle.

Use this skill when creating new Codex skills from scratch, editing existing skills to improve their descriptions or structure, or converting Codex sub-agents to skills. This includes designing skill workflows, writing SKILL.md files, organizing supporting files with intention-revealing names, and leveraging CLI tools and Node.js scripting.

Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Codex's capabilities with specialized knowledge, workflows, or tool integrations.