Standard tier

The default for serious teams. Full six-role workflow, per-role MCP scoping, change-summary discipline, conformance verification. The tier most of this wiki implicitly documents.

Audience #

Pick Standard if any of these describe you:

You've outgrown a hand-rolled CLAUDE.md and want explicit roles you can read, edit, and reason about.
You run more than one project with the same AI coding CLI and want a reproducible scaffold rather than copying a folder around.
You're a 2–10 person team that has felt scope-leak or context-loss pain and wants enforcement, not just convention.
You don't yet need audit artifacts but do want change summaries on every task so your PRs are readable.

Install #

npx aidokit init --tier standard

Standard is also the default for non-interactive use; aidokit init --yes gives you this tier unless you say otherwise. The interactive picker offers it as the recommended choice.

What you get #

Component	Standard
`CLAUDE.md`	Full project rules with source-of-truth chain and anti-hallucination guards
Slash commands	`/intake`, `/implement-task`, `/orchestrator-next`
Roles	Six explicit roles + the optional frontend browser tester
Per-role MCP scoping	Yes — Researcher gets Context7 / Beads MCP; Builder does not
Task graph	Beads required (with prereq-check hint to install)
MCP servers	Suggested by your stack pack; you pick during init
Watchdog hooks	Scope check blocks out-of-scope writes; deletion guard; version guard
`docs/` skeleton	Full set: brief, architecture summary, decisions, testing, research
Change-summary discipline	Every `/implement-task` produces an artifact under `agent-artifacts/changes/`
Validation commands	Wired into the watchdog so the agent runs them before declaring done
Conformance verification	`aidokit doctor` reports drift against your declared tier

What you do not get (and why) #

Capability declarations and signed manifests. These are Strict tier features and require additional artifact discipline that Standard doesn't impose. Most teams don't need them.
Audit export commands. aidokit audit export --format soc2 and --format eu-ai-act are Strict-only. They build on top of the capability declarations.
Strict-mode policy-as-code. A future Strict-only feature for declarative agent permissions.

Coming from Starter? #

aidokit sync --tier standard --dry-run    # preview
aidokit sync --tier standard               # apply

Your task graph, decisions, and previous artifacts are preserved. The engine directory is augmented in place; CLAUDE.md is updated only in sections marked managed-by-aidokit.

Coming from a hand-rolled CLAUDE.md? #

Use brownfield mode to ingest what you already have:

aidokit init --tier standard --brownfield

aidokit will scan for existing CLAUDE.md, docs/, and any PRD/BRD-shaped documents and offer to fold them into the Standard layout. Nothing is overwritten without a diff preview.

Honest limitations of Standard #

The full six-role vocabulary is real cognitive load. The first project feels heavy; subsequent projects compound the benefit.
MCP scoping requires you to think about which agent needs which tool. The defaults are good, but you'll want to refine over time.
Conformance verification is local-only; CI doesn't yet check this automatically (planned for v1.x).

Upgrade to Strict #

When Standard isn't enough (regulated industry, audit requirement, policy-as-code needs, signed package supply chain), upgrade:

aidokit sync --tier strict --dry-run
aidokit sync --tier strict

Be ready for new artifacts: capability declarations, signed manifests, audit exports. Strict has higher discipline cost; see Strict tier before deciding.