`core` — guides

core is the flagship pack: one front door for durable work intake, then a loop your agent can’t cut corners in. Describe the work in ordinary language; work-intake routes it to the right artifact and lifecycle state. Approved specs continue through hard gates and cold review.

Start work on adding export retention controls for workspace owners.

New here? Why loop engineering is the why — the leverage has moved off the prompt and onto the loop. The core pack as a system is the full map. Then build something with plan and execute non-trivial work.

Tutorials

Learning-oriented, start-to-finish.

• From idea to a walking skeleton — stand up a new project the structured way, end to end.

How-to

Task-oriented recipes for a problem you already have.

• Plan and execute non-trivial work — the loop itself, applied to a feature or change.
• Start or remember work without choosing a skill — route an ordinary request into the right artifact and workspace state.
• Fix a bug — the diagnose-then-fix path, with a regression test as the receipt.
• Adapt a freshly-installed pack to your project — tailor the defaults to your repo after install.
• Review a branch or PR you didn’t write — point the reviewers at anyone’s diff.
• Close work without losing lasting context — verify durable context, settle coordination, and choose a safe disposition.
• Intake an external brief into a product brief — turn unstructured external input (email, message, issue) into a DoR-ready product brief.
• Receive a product brief and decompose it into specs — turn a multi-feature handoff into shippable specs.
• Run a 30-minute live workflow demo — show technical, enterprise, or non-technical teams the canonical Core and Product Engineering handoffs on their own repository.
• Decide and record your foundation during inception — the ADR + reference.md you write before the first feature.
• Migrate a legacy workspace entry safely — convert one reviewed compatibility entry with ledger-backed recovery and rollback.

Reference

Information-oriented, dry and complete.

• Spec Shape: and the plan’s ## Design (LLD) — the fields, what they mean, and how the stack is derived.
• Product brief fields — the brief field list and the linkage it stamps on derived specs.
• Work-intake routing and lifecycle — exact routes, states, processors, and mutation boundaries.
• workspace.toml schema reference — target entries, compatibility forms, findings, migration policy, and ledger state.

Explanation

Understanding-oriented — the why behind the design.

• The core pack as a system — how the parts compose, and how it compares to vibe-coding, Spec Kit, and Kiro’s spec mode.
• The token economy of the loop — what the loop wastes, what it spends on purpose, and why the cold reviewer earns its cost.
• Why the plan owns the LLD — where the low-level design lives and why it isn’t in the spec.
• About the walking skeleton — the thinnest end-to-end slice, and when to throw code away instead.
• About foundation vs. map — the two things you record at inception and why they’re different.
• Why a brief layer — why a brief sits above the spec when work spans many features.
• Why work begins with an artifact — how source, durable meaning, lifecycle, and execution remain separate.
• Role journeys — how PMs, engineers, and agents use the system at their operating altitude.

---

Cross-cutting guides — installing the catalogue, upgrading packs, the adapter support matrix — live in ../_shared/.