AgentsKit — Master Execution Plan

Master execution plan for the 149 issues (Phase 0 + Phases 1–4).
Sequence, dependencies, sprints, gates, and suggested order.

Related documents:

• Master PRD #113 — 97 product roadmap issues
• Phase 0 PRD #211 — 52 foundation hardening issues
• PHASE-0-EXECUTION-PLAN.md — detailed P0→P2 plan
• Project Board

Overview

Total: ~12 months of a clear roadmap. Mandatory quarterly review to recalibrate.

Sprint Structure

2-week sprints. Over 12 months: ~26 sprints.

Phase → Sprint → Issues Mapping

Note: assumed capacity = 1–2 devs. With more people, compress sprints.

Critical Dependencies

Before starting Phase 1

• ✅ Contract ADRs ([P0.3] ADRs for 6 core contracts (Adapter/Tool/Memory/Retriever/Skill/Runtime) #214) completed
• ✅ Bundle/coverage gates ([P0.6] Bundle size budget enforced in CI (size-limit) #217 [P0.7] Coverage gate in CI (thresholds per package) #218) active
• ✅ agentskit.io live ([P0.24] Configure agentskit.io (DNS + SSL + redirect) #235)
• ✅ Fumadocs migrated ([P0.27] Decision Tree (5 questions → packages) #238)
• ✅ README + Manifesto + Origin ([P0.13] Rewritten root README (tagline + origin + before/after + honest comparison) #224 [P0.14] Public MANIFESTO.md (non-negotiable principles) #225 [P0.15] ORIGIN.md (why it exists, author's pain) #226) published
• ✅ Playwright E2E in examples ([P0.40] E2E Playwright on 4 apps/example-* #251)

Feature dependencies

#3 Contract ADRs ─────┬──> entire Phase 1 (needs stable contracts)
                      └──> entire Phase 3 (new adapters/tools)

#54 MCP bridge (P3) ──> #31 agentskit ai (P2) — needs MCP tools
#21 Deterministic replay (P2) ──> #39 Replay sessions + #43 Test fixtures
#77 AgentsKit Edge (P3) ──> #86 Cloud free tier (P4) — edge runtime
#69 Skill marketplace (P3) ──> #91 Revenue share (P4)
#81 A2A Protocol (P3) ──> strategic partnerships (P4)

Capacity warnings

• S16–S18 (memory + security + durability) are heavy — consider 3 sprints instead of 2
• S19 major adapters requires API keys for 10+ providers — prepare budget/credentials
• S20 MCP bridge is architecturally heavy — assign a senior dev

Phase Gates (Do Not Skip)

Gate Phase 0 → Phase 1

• 6 core ADRs merged
• Bundle/coverage gates blocking PRs
• agentskit.io resolving + docs.agentskit.io live
• Root README rewritten + Manifesto + Origin published
• Fumadocs with Concepts section + ≥5 recipes
• Migration guides (LangChain + Vercel AI) published
• Playwright E2E running in 4 examples
• Discord with ≥20 members + CODEOWNERS active
• LICENSING.md decided
• Coordinated launch executed (HN/Twitter/PH/Reddit)

Gate Phase 1 → Phase 2

• 20 Phase 1 stories done or explicitly descoped
• npx agentskit init with ≥1000 downloads/month
• Discord with ≥200 members
• ≥3 external users sharing public projects

Gate Phase 2 → Phase 3

• Deterministic replay + working devtools (real differentiation)
• Complete security layer
• Hierarchical memory + RAG reranking shipped
• ≥10 stars/week consistent growth
• ≥1 published case study

Gate Phase 3 → Phase 4

• 10+ stable adapters
• Bidirectional MCP bridge working
• ≥1 strategic partnership announced (Vercel, Cloudflare, provider)
• User base ≥5k DAU

Rhythm & Ceremonies

Daily

• Status update on Discord #maintainers (2 lines — done/doing/blocked)

Weekly — every Friday 5 PM

• Sprint Review (30 min): demo completed work
• Planning (30 min): adjust priorities

Biweekly — end of each sprint

• Retro (30 min): what worked, what didn’t, what to change
• Update Project board: move completed, re-estimate pending

Monthly

• Community update — blog/newsletter
• Full roadmap review — does any phase need recalibration?

Quarterly

• Public roadmap review — RFC for adjustments
• Metrics evaluation — stars, downloads, Discord, contributors

Success Metrics

Phase 0 (3 months)

• agentskit.io indexed on Google
• Coordinated launch with ≥1000 visits in the first week
• ≥500 GitHub stars
• ≥20 Discord members

Phase 1 (end of month 4)

• ≥1000 npm downloads/month
• ≥200 Discord members
• ≥5 external contributors with merged PRs

Phase 2 (end of month 7)

• ≥5000 downloads/month
• ≥1 Hacker News front page post
• Coverage in ≥3 tech newsletters (Node Weekly, JS Weekly, etc.)

Phase 3 (end of month 10)

• ≥20k downloads/month
• 10+ mature adapters
• ≥10 public open-source projects using AgentsKit

Phase 4 (end of year 1)

• Cloud with ≥100 active users
• First enterprise customer
• Infrastructure cost break-even

Project Board — Recommended Setup

Add Iteration field (sprint tracker):

• Options: S1 → S27
• Pre-fill for next 2–3 iterations

Create Views:

1. 🔥 Current Sprint — filter: iteration = current, group by Status
2. 📅 Roadmap Timeline — layout: roadmap, group by Phase
3. 🎯 By Priority — filter: current/next, group by Priority
4. 🧭 Backlog Planning — filter: empty iteration, group by Phase
5. 📦 By Package — group by Category/Package

Next Concrete Actions (Start Today)

This week (Sprint 1 start)

1. Create branch foundation/manifesto-origin — draft MANIFESTO.md + ORIGIN.md
2. Buy DNS service + Cloudflare for agentskit.io — point temporary CNAME
3. Setup size-limit in a test package (core) — validate flow before scaling
4. Create issue/PR templates in .github/ — unblock future work
5. Decide doc host: Vercel vs Cloudflare Pages (recommend Vercel for Fumadocs/Next.js integration)

Next week

1. Finalize MANIFESTO + ORIGIN (approved PR)
2. Start rewriting root README (block by block)
3. Fumadocs spike in separate branch
4. Open first ADR (0001-adapter-contract.md) as public RFC

Week 3

1. Docusaurus vs Fumadocs parity check
2. Continue ADRs (goal: 1/day)
3. CODEOWNERS + issue templates merged
4. Enable bundle/coverage gates on main

Execution Principles

1. Done is better than perfect — ship v0.1 and iterate
2. Build in public — RFCs, ADRs, decisions all open
3. Aggressive time-boxing — every task has a limit; if exceeded, reassess
4. Move at least 1 issue per day — momentum matters
5. Never start a phase with the previous incomplete — gates exist for a reason
6. Dogfood — use AgentsKit to build AgentsKit (docs chat, release agent, etc.)