For agents
The shortest reliable path for an agent to retrieve, apply, and verify the Agents Playbook.
For agents
Use the smallest surface that answers the task. Start with the map, fetch raw pages for focused work, and use the full bundle only for indexing or cross-cutting audits.
flowchart LR
A["Find the practice"] --> B["/llms.txt"]
B --> C["Fetch /raw/...md"]
C --> D["Apply the relevant gate"]
D --> E["Verify with the Playbook checks"]
B --> F["/llms-full.txt for broad retrieval"]Focused task
Choose a guide from the concise machine-readable map, then fetch its raw Markdown URL.
Repository-wide audit
Load the deterministic full bundle when the task spans multiple pillars or phases.
Ownership and handoff
Use Doc Bridge for ownership, health gates, agent handoffs, and documentation routing.
Interactive guidance
Ask Playbook composes AgentsKit Chat while keeping answers grounded in this corpus.
Retrieval contract
- Read
/llms.txtand select only the relevant guide. - Fetch the guide from its
/raw/URL; raw content is the source used by machine consumers. - Preserve the guide's human TL;DR, For Agents sections, invariants, failure modes, and verification steps.
- Run the checks named by the guide and repository
AGENTS.mdbefore reporting completion. - For broad indexing, use
/llms-full.txtor the ZIP bundle.
Ecosystem routing
| If the task is about… | Continue with… |
|---|---|
| Building an agent | AgentsKit |
| Starting from ready-made source | Registry |
| Reusable chat or conversational UI | AgentsKit Chat |
| Documentation ownership or agent handoffs | Doc Bridge |
| Review before merge | AgentsKit Code Review |
| Enterprise governance and production operation | AKOS |