Agents Stack Starter

This scaffold is a file-backed agents-stack project.

If you are already inside the generated scaffold, skip the copy step and run ./docs/scripts/init.sh.

# From the starter source, copy the scaffold into a clean project directory.
cp -R templates/. my-project
cd my-project

# Create any missing baseline files and directories without overwriting work.
./docs/scripts/init.sh

If you want to clone the starter from GitHub instead of copying locally:

npx degit thompson0012/agents-stack/templates new-project && cd new-project && ./docs/scripts/init.sh

After that, the repo is ready for truthful initialization work.

The copied scaffold starts blank: no active sprint, no archive history, and no seeded feature backlog.

What to edit first

1. Set the project name in docs/live/tracked-work.json.
2. Record the real source goal in docs/live/roadmap.md and docs/live/current-focus.md.
3. Add the first backlog item only when it is real, bounded, and ready to track.
4. Fill in docs/reference/architecture.md and docs/reference/design.md with the truth of the new project.

What lives where

• AGENTS.md — the repo constitution and harness rules.
• docs/live/tracked-work.json — the authoritative backlog and single runnable-sprint selector.
• docs/live/ideas.md — pre-sprint exploration and rejected directions.
• docs/live/progress.md — reviewed outcomes and durable next actions.
• docs/live/memory.md — durable lessons captured after compounding.
• docs/reference/* — stable architecture and design intent.
• .harness/<feature-id>/ — sprint-local execution state.
• docs/archive/* — finished sprint history.

Normal harness flow

project-initializer -> generator-brainstorm -> generator-proposal -> evaluator-contract-review -> generator-execution -> adversarial-live-review -> state-update -> compound-capture

Keep exactly one runnable sprint at a time. If docs/live/tracked-work.json is missing or untrustworthy, start with project-initializer.

Notes

• ./docs/scripts/init.sh is safe to run repeatedly; it only creates missing baseline files.
• docs/live/tracked-work.json is the source of truth for what is runnable right now.
• docs/live/ideas.md is not a runnable schedule.