# Roadmap: claude-howto 2026–2027 > April 2026 – March 2027 · Dual-Layer Knowledge Base | Full plan: `TASKS-20260401.md` --- ## Vision Transform claude-howto from a static tutorial repo into a **living, dual-audience knowledge system**: - **For humans** — interactive, scenario-based learning with progressive difficulty, decision trees, and named patterns that experts bookmark - **For AI agents** — structured metadata index that agents query before performing Claude Code tasks, making this repo infrastructure, not just content No competitor targets AI agents as a primary audience. This is the moat. --- ## The 7 Pillars | # | Pillar | What it delivers | |---|--------|-----------------| | P1 | Fun Layer | Scenario intros + "Try It Now" blocks in every module | | P2 | AI Agent Index | Generated `agent-manifest.json` + `AGENT-INDEX.md` + lookup skill | | P3a | Expert Reference (in-module) | Decision trees + named patterns per module | | P3b | Expert Reference (cross-module) | `RECIPES.md` — compound multi-feature workflows | | P4 | Newcomer Onboarding | `quickstart.sh` + `QUICKSTART.md` + difficulty badges | | P5 | Community Showcase | `COMMUNITY-PROJECTS.md` — curated user projects | | P6 | Content Quality | Expand weakest modules; project `CLAUDE.md` | | P7 | Living Curriculum | `WHATS-NEW.md` + version badges + weekly staleness CI action | --- ## Timeline at a Glance ``` Apr 2026 May–Jun 2026 Jul–Aug 2026 Sep 2026 Oct–Nov 2026 Dec 2026–Mar 2027 | | | | | | [M1] [M2] [M3] [M4] [M5] [M6] Infrastructure 6/10 modules 10/10 Agent layer Version audit Self-sustaining + hooks/checks complete complete + recipes complete system ``` --- ## Milestones ### M1 — Infrastructure Live · End of April 2026 **What ships:** - `scripts/quickstart.sh` — one-command setup for new users (idempotent) - `QUICKSTART.md` — First 15 Minutes visual guide - Difficulty badges + "What you'll build" previews on all 10 modules - `WHATS-NEW.md` + version badges on all modules - `.github/workflows/staleness-check.yml` — weekly Monday issue if module unverified 30+ days - Root `CLAUDE.md` — project's own configuration as a best-practice example - `scripts/build-agent-index.py` — generator that reads all 10 modules → outputs `agent-manifest.json` + `AGENT-INDEX.md` - **06-hooks** — full deep pass: 5 complete hook scripts, decision tree, Try It Now, patterns - **08-checkpoints** — full deep pass: 311 → 800+ lines, 3 workflow templates, decision tree, patterns **Why start here:** Infrastructure benefits every future phase. Hooks and checkpoints are the weakest modules and most likely to lose new visitors. --- ### M2 — 6/10 Modules Complete · End of June 2026 **What ships (one deep pass per module):** - **01-slash-commands** — scenario intro, decision tree, Try It Now, named patterns - **02-memory** — scenario intro, decision tree, Try It Now, named patterns - **03-skills** — scenario intro, decision tree, Try It Now, named patterns - **10-cli** — scenario intro, decision tree, Try It Now, named patterns - CI step: validate `agent-manifest.json` schema on every push Each module pass = run the generator to confirm valid manifest output. --- ### M3 — All 10 Modules Complete · End of August 2026 **What ships:** - **04-subagents** — full deep pass (incl. "The Multi-Agent Review Pattern") - **05-mcp** — full deep pass - **07-plugins** — full deep pass - **09-advanced-features** — full deep pass Every module now has: scenario intro, 2+ Try It Now blocks, Mermaid decision tree, 2+ named patterns. --- ### M4 — Agent Layer Live · End of September 2026 **What ships:** - Final `agent-manifest.json` covering 100% of modules (generated from completed content) - `AGENT-INDEX.md` linked from `README.md` - `skills/claude-howto-lookup/SKILL.md` — lightweight agent skill that queries the manifest - `RECIPES.md` — 5+ compound workflows (schema: name, modules-used, problem, solution, expected outcome) - `COMMUNITY-PROJECTS.md` — static curated list with PR-based submission format **Why September:** The agent index is only meaningful once all 10 modules are content-complete. --- ### M5 — Version Audit Complete · End of November 2026 **What ships:** - Full version audit: all 10 modules verified against current CC version - Updated `cc_version_verified` frontmatter + version badges everywhere - `RECIPES.md` expanded to 8+ recipes based on observed community patterns - Pinned GitHub Discussion: "Share your Claude Code workflows" — signal collection for agent usage --- ### M6 — Self-Sustaining System · End of March 2027 **What ships / runs continuously:** - `/docs-sync-claude-code` skill runs after every CC release → `WHATS-NEW.md` updated - Agent manifest CI regression: 100% module coverage enforced - `RECIPES.md` at 10+ recipes - `COMMUNITY-PROJECTS.md` growing organically - Agent usage signal evaluated → if validated, promote the lookup skill (marketing, asm registry) --- ## Deliverables Summary | Deliverable | Type | Phase | |-------------|------|-------| | `scripts/quickstart.sh` | Script | P1 | | `QUICKSTART.md` | Doc | P1 | | Root `CLAUDE.md` | Config | P1 | | `WHATS-NEW.md` | Doc | P1 | | `.github/workflows/staleness-check.yml` | CI | P1 | | `scripts/build-agent-index.py` | Script | P1 | | 10 module deep passes (scenario + Try It Now + decision tree + patterns) | Content | P1–P3 | | `agent-manifest.json` (generated) | Data | P4 | | `AGENT-INDEX.md` (generated) | Doc | P4 | | `skills/claude-howto-lookup/SKILL.md` | Skill | P4 | | `RECIPES.md` (5 → 8 → 10+ recipes) | Doc | P4–P6 | | `COMMUNITY-PROJECTS.md` | Doc | P4 | --- ## What Is NOT in Scope Deferred to `TODOS.md` — do not let these creep in: - Skill marketplace or installable registry - Custom website or dashboard - Completion tracking (cc-progress) - Community tutorial CI validation - Auto-generated CONTRIBUTORS.md - Multi-language translations - Quiz/assessment infrastructure - Community voting on projects