AI-DLC KitThe pinned AWS AI-DLC methodology, run as a Flow Agents kit on Flow gates and Surface evidence

AI-DLC Kit#

The independently maintained AI-DLC methodology kit for Flow Agents. It uses the same public kit installation, activation, Flow gate and host adapter contracts as any other kit. Upstream methodology and license provenance are frozen under upstream/.

Install the Git repository at an immutable revision into an initialized Flow Agents workspace:

flow-agents kit install https://github.com/kontourai/aidlc-kit.git#<commit> --dest /path/to/workspace
flow-agents kit activate --dest /path/to/workspace --format json

The npm package is also a kit repository: unpack its tarball and pass the resulting package directory to flow-agents kit install. Installation copies a validated repository; it does not fetch or execute npm dependencies. Runtime applications install this package normally to resolve its pinned public Flow and Surface dependencies.

npm ci
npm run check
npm test
npm run test:package

Package-install verification packs this repository and performs a clean npm installation outside its source tree. It validates and activates the installed kit through the public Flow Agents CLI and exercises the packaged public Flow/Surface contracts. Live model equivalence is a separate measurement; no synthetic package check proves effectiveness.

Documentation is published to https://kontourai.github.io/aidlc-kit/ with docs-kit. docs-kit.json lists the published docs and the topics served to agents, and npm run docs:check is the gate. When you change code that a document covers, update that document.

See execution and methodology for controller, authority and evidence boundaries.

For ordinary installed execution, install this repository through npm in the destination workspace, then install and activate that directory as a kit. The kit pins the published Flow Agents 7.1.0 runtime, including its public /kit-execution, /docker-worker and /provider-broker exports. The kit itself is distributed from Git; select an immutable reviewed commit for reproducible installation.

npm install "git+https://github.com/kontourai/aidlc-kit.git#<reviewed-commit-sha>"
flow-agents kit install node_modules/@kontourai/aidlc-kit --dest .
flow-agents kit activate --dest . --format json
flow-agents kit run aidlc --dest . --request /private/request.json --controller-root /private/run-controller --auth-file /private/codex-auth.json

The ordinary request uses schema: "kontour.kit.execution_request", version: "1.0", kit_id: "aidlc", a task prompt, run_id, absolute model workspace, finite execution model/turn/time settings, and engine_sandbox.image as an immutable Docker image ID. Kit runtime dependencies resolve in the destination workspace; the model workspace is separate. Host prerequisites are Node 24, Docker (Desktop on macOS or Engine on Linux) and Python 3 for descriptor-relative storage inspection. Keep controller state, credentials and hidden grading materials outside worker mounts.

For a Linux host with an incoming firewall, configure engine_sandbox.provider_broker with an explicit IP bind_host and TCP port. The public worker maps host.docker.internal to the Docker host gateway on Linux. Permit only the relevant Docker bridge source range to reach that listener; the scoped provider capability remains mandatory. The default uses an ephemeral port, and this configuration never disables or changes the host firewall. Configure reachability before starting a run; a connection failure is retained rather than counted as provider work.

{"engine_sandbox":{"provider_broker":{"bind_host":"172.17.0.1","port":8326}}}

Use the actual bridge gateway and an available port on your host. Broker transport settings are part of the run binding; changing them starts a new registration rather than reusing prior approvals or provider quota.

For native login, Datum provider references, Docker broker setup, engine qualification and ordinary resume commands, read Environment setup.

Documentation

Start Here

Environment setupThis kit runs AI-DLC through Flow Agents and Flow. It does not invoke the AWS aidlc engine. Use flow-agents kit run aidlc to avoid confusing the two products' CLI names. The package exposes kontour-a…

Concepts

AI-DLC Reference KitAWS AI-DLC's pinned 33-stage methodology and 11 profiles, expressed through Kontour's existing Flow definitions and Flow Agents kit contracts. This independently installable reference exercises the s… Pinned AI-DLC methodologyThe AI-DLC reference kit imports the methodology of AWS AI-DLC Workflows source at commit 2a555f7a4387cd1b8254029cccb9fd1e99974574. The imported material is licensed MIT-0; its original license is re… Execution and evidence boundariesFlow owns the canonical run, gate evaluations, route-back and retry budgets. Flow Agents owns kit bindings, skills and host execution. Veritas owns repo requirements and repository readiness. Surface…

Guides

Reply to questions through the ordinary operator inboxdecide --disposition accepts explicit questionnaire responses. Approval by itself is not an answer. decisions lists the signed pending input, including each questionnaire document, its stable answer-… Request changes after an advisory reviewUse the ordinary signed decision inbox to request a correction to the current reviewed work. Supply exact current unit and open finding IDs and the operator's clarification in a private disposition f… Change pending stages in a running workflowLive recomposition changes the pending stage grid of an existing run. It does not change the original request, its selected scope/depth/testing policy, completed work, current execution contract, or… Revising an observed public inputUse the ordinary kontour-aidlc revise-input command when an operator changes an already observed public input while execution is stopped at its actual signed wait. This is separate from requesting ch…

Reference

ReferenceThis page lists the public command and module surface of @kontourai/aidlc-kit. Ordinary execution goes through flow-agents kit run aidlc; see Environment setup. The kontour-aidlc command covers metho… Standard kit execution entryA kit declares a standard entry in kit.json:

Source: README.md