📖 Browse the patterns as a human-readable site: ramparte.github.io/agent-building-playbook
A curated, growing collection of patterns for building better in the agentic age — Programming Pearls, but actionable and for the cognitive-architecture era.
The subject is how to build better, not narrowly how to build agents. Patterns here cover cognitive architecture, agentic workflow design, prompt engineering, orchestration, tooling, and the everyday craft of working productively alongside AI systems.
One canonical source (markdown) serves two audiences:
- Agents read INDEX.md — a dense, token-efficient list with one-line descriptions grouped by tag, each linking to its full file.
- Humans browse the
patterns/directory directly on GitHub or a generated wiki, reading full pattern pages at their leisure.
| Path | Description |
|---|---|
INDEX.md |
Dense agent-facing index. Generated — do not hand-edit. |
SUBMISSIONS.md |
How to write a pattern and open a PR. |
patterns/ |
One file per pattern: flat frontmatter + markdown pearl body. |
scripts/build-index.sh |
Regenerates INDEX.md from patterns/. |
- Read SUBMISSIONS.md.
- Add
patterns/{kebab-title}.md. - Open a PR — don't edit
INDEX.md(it is generated).
scripts/build-index.sh reads every patterns/*.md and rewrites INDEX.md:
bash scripts/build-index.shA GitHub Action runs the same script on every merge to main, so INDEX.md is always up-to-date in the repo.
Patterns here are distilled from three sources: external literature, battle-tested session history, and the Amplifier team ecosystem. Personal lessons and external literature are treated as peers — there is no provenance weighting in v1. The original research dump that seeded this collection is preserved at docs/seed-research-dump.md.