Workflows20 workflows in 13 sections
Workflows
End-to-end playbooks for shipping real software with AI agents. Not prompts, not tips — repeatable processes that senior teams can drop into an existing project and run this week.
Context Engineering
Open the section: Context EngineeringPrompt engineering is dead. Context engineering is what actually makes AI agents useful on real projects. It's the practice of architecting what your agent knows before it writes a single line — so it behaves like a senior engineer on your team, not a tourist.
- 2 min readCLAUDE.md That Actually WorksStop re-explaining your project to every new agent session.Read
- 4 min readAGENTS.md — One File, Every AgentThe open standard that makes your project context portable across Claude Code, Cursor, Codex, Copilot, and 20+ other agents. Write it once, every agent reads it.Read
- 5 min readWriting Your Own SkillsThis hub is full of skills other people wrote. The compounding win is writing your own. A skill is how you encode your team's process once and have every agent follow it the same way.Read
- 1 min readExplore → Plan → ImplementSeparate research from implementation so you stop solving the wrong problem.Read
- 1 min readWriter / Reviewer PatternA second Claude session reviewing code it didn't write catches more than a self-review.Read
Design Systems for Agents
Open the section: Design Systems for AgentsEverything above is language-agnostic. This section is the frontend one, because frontend has a failure mode the general playbooks do not catch: output that passes every check you run and is still off-system.
Long-Running Sessions
Open the section: Long-Running SessionsContext engineering decides what the agent knows when the session starts. This section is about hour three, when the window is full, the conversation gets summarized, and half of what you carefully loaded quietly stops being there.
Planning & Scoping
Open the section: Planning & ScopingVerification
Open the section: VerificationPlanning tells the agent what to build. This section is about how it proves the work is right, so that "done" is a result the agent measured and not a claim it made.
- 6 min readThe Browser Verification LoopThe highest-leverage change a Frontend team can make to its agent setup. Until the agent can open the page it just built, "done" only ever means "it compiled", and you are the verification loop.Read
- 9 min readGolden-PR EvalsEvery team has an opinion about whether their agent setup is any good. Almost nobody has a number. An eval suite built from your own merged pull requests turns "the agent is doing great" into a score you can regress against, and it is the only honest way to compare a model, a CLAUDE.md rewrite or a new skill.Read
- 6 min readEvals for the LLM Features You ShipA feature that calls a model does not give the same answer twice, so a green unit test says very little about it. The eval set that protects it cannot be written in a planning doc. It gets harvested from what the feature actually did to real users.Read
Review & Merge
Open the section: Review & MergeVerification tells you the change is right. This section is about the last hundred metres, where a change that is right still has to become a change that is merged, and where most of what an agent produced in 2026 quietly stopped.
Migrations at Scale
Open the section: Migrations at ScaleEvery other playbook here is about producing code. This one is about changing code that already exists, thousands of files at a time, without a rewrite, without a freeze, and without a quarter of your team reading diffs.
- 10 min readCodemod-First MigrationsA migration is not one hard problem. It is one easy problem four hundred times, and an agent is the worst tool you own for the four hundredth repetition. Stop asking it to perform the change. Ask it to write the change, then run that.Read
- 9 min readFleet MigrationsOnce the codemod has eaten the mechanical work, what remains is too varied for a transform and too large for one session. That residue is the job for a fleet. The hard part is not spawning agents. It is cutting units that can each merge alone.Read
Agent Architecture
Open the section: Agent ArchitectureParallel Execution
Open the section: Parallel ExecutionAutonomy & Isolation
Open the section: Autonomy & IsolationVerification tells you the work is right. This section is about how far the agent is allowed to run before a human looks at it, and what contains the damage on the run where something goes wrong.
Supply Chain & Trust
Open the section: Supply Chain & TrustEvery section above hands your agent more context and more autonomy. This one is the bill that arrives with it. The files you added to the repository so the agent would read them are now the highest-value place in your codebase to put an attack, they arrive through the same registry your components do, and nothing in your pipeline has ever read them.
Distribution & Governance
Open the section: Distribution & GovernanceEverything above makes one engineer's agent good. This section is about making every engineer's agent good, and keeping it that way after the person who set it up moves on.
Continuous AI
Open the section: Continuous AIEverything above runs because a person opened a session. This section is about the work that runs when nobody did.