adr-coveragelisted
Install: claude install-skill real-case/marvin-toolkit
# ADR Coverage
Gap analysis between the **recorded** decisions and the **actual** ones. Every codebase is full
of decisions — a framework, a database, a protocol, a build pipeline — and only some of them are
written down. This command surfaces the significant undocumented ones and ranks them, so the
team can pay the documentation debt highest-blast-radius first.
Strictly read-only: no records are created, no files written, no tool mutation. The output is a
ranked candidate list in chat, each entry ready to hand to `/marvin:adr`.
## Input
`$ARGUMENTS` — optional focus area (e.g. "infra", "the storage layer", "frontend choices").
The analysis still spans the project; a focus deepens that area.
## Workflow
### Phase 1 — What is on record
Call the `adr` MCP tool with `{"action": "list"}` and build the coverage map: for each record,
the decision area it covers (storage, transport, build, testing, deployment, architecture
patterns, process). Read titles first; open records only where the title is too vague to
classify. Note explicitly rejected or superseded decisions — those areas *are* covered.
### Phase 2 — What the stack actually decided
Ground the map against the repository. Sweep the decision-bearing surfaces:
- **Dependencies** — manifests (`package.json`, `go.mod`, `requirements.txt`/`pyproject.toml`,
`Cargo.toml`, `pom.xml`, …): the framework, the ORM/DB driver, the test runner, the state
or messaging libraries. A major dependency is a decision someone made.
-