Docs / Configuration

Configuration

Project config — .asdt/config.yaml

Running /asdt-init (inside your assistant) creates .asdt/config.yaml with sensible defaults. Edit it to match your project.

memory:
  provider: engram

specialists:
  strict_tdd: false

memory.provider

The memory backend used by all specialists to persist and retrieve artifacts. Only engram is supported today.

specialists.strict_tdd

When true, the Developer specialist’s workflow includes a test step after implement. Tests are produced as structured artifacts — not executed — so this flag controls generation, not CI.

Default: false.

Platform context — .asdt/knowledge/platform.yaml

Generated by /asdt-init. Describes the technical context of the project so every specialist can tailor its output without re-detecting the stack.

language: typescript
framework: astro
package_manager: npm
test_runner: vitest
lint: eslint

Fields are free-form — specialists read what’s present and skip what’s missing. You can add custom fields like deployment: vercel or database: postgres and reference them in specialist prompts via the platform-context shared skill. That fragment is not ambient — it loads only where a specialist declares it, either as an inline platform-analysis step in workflow.yaml or in a step’s reference_skills: list.

Environment variables

ASDT reads no environment variables directly. The binary uses the AI assistant’s environment (Claude Code or OpenCode) for all runtime configuration.

Multiple projects

Each project has its own .asdt/ directory. Switch projects by opening your AI assistant in a different directory — ASDT reads configuration from the nearest .asdt/config.yaml walking up from the current working directory.