Rebar Onboarding Walkthrough

patterns rebar onboarding

Standard walkthrough for onboarding a new team member to the Rebar framework. Covers the four knowledge systems, day-to-day commands, the self-learn loop, and the close-loop harness.

Key Points for New Engineers

1. Start with client.yaml — copy from clients/_templates/client.yaml, fill in, run /discover
2. Phase 0 — auto-generated tech stack analysis, architecture assessment, potential gotchas
3. Day-to-day — /brief for standups, /check for compliance, /improve after investigations
4. After a feature ships — run /close-loop <feature> to trigger the four-gate harness (evaluator → release gate → /improve —from → /meta-improve queue → /wiki-ingest)
5. Wiki intake — drop files in raw/, run /wiki-ingest
6. Compounding loop — good answers from questions get filed via /wiki-file <topic>
7. Template changes — /meta-improve queues patches; run /meta-apply in main session to review
8. Maintenance — /wiki-lint periodically for orphans, broken links, stale pages

Why Not Just a Wiki?

Four systems exist because of different access patterns:

• YAML — machine-readable, parsed programmatically by commands
• Memory — injected silently into every session, behavioral rules
• Skills — tactical playbooks auto-discovered by Claude Code’s skill router
• Wiki — cross-linked, human-readable, cross-project

Mixing them creates something hard to query and hard to maintain.

The “Aha” Moment

“The LLM reads index.md first to navigate, follows the wiki links, and synthesizes answers with full context. No searching, no stale docs that nobody updates. The system maintains itself as a side effect of doing work.”

Related

• three-systems — detailed explanation of the three knowledge layers
• self-learn-loop — how expertise.yaml accumulates knowledge
• commands — which commands update which system

---

Source: raw/example-transcript.md | Ingested: 2026-04-13