Mechanical rename produced by scripts/msd-rename.cjs: gsd/Gsd/GSD -> msd/Msd/MSD across contents and paths, upstream package/repo coordinates -> @golem15/msd-core and golem15com/msd-core. Deep links into upstream history, sibling upstream packages, the GSD-2 import feature, CHANGELOG.md and .changeset/ are kept as-is. Hand edits on top: MSD block-letter banner and logos, LICENSE copyright line, package/plugin identity, regenerated lockfile, install-tree fixtures, derived registries and benchmark baseline; migration checksum baseline re-locked (MSD keeps its own install state, so no install had applied the old sums); sort-order and regex-escaped expectations in tests adjusted.
9.3 KiB
Existing Code Onboarding Module owns deterministic repo-state detection and onboarding route selection
- Status: Accepted — ratified 2026-07-17 (originally Proposed 2026-07-06); see "Ratification" below
- Date: 2026-07-06
- Issue: #1990
- Implementation: PR #1994
Ratification (2026-07-17): Proposed → Accepted
Ratified by explicit maintainer directive after independent re-verification of the evidence below; the Status field sat at Proposed for 11 days after the decision shipped.
Evidence the decision shipped:
- Issue #1990 ("Add /msd:onboard for existing-codebase setup") is CLOSED, stateReason COMPLETED, closed 2026-07-07T04:23:56Z; PR #1994 ("feat(#1990): add brownfield onboarding workflow") is MERGED into
nextat 2026-07-07T04:23:55Z with bodyCloses #1990. src/onboard-projection.cts(15,248 bytes) and its compiledmsd-core/bin/lib/onboard-projection.cjsare both present on disk, implementing the projection this ADR describes.src/init.cts:56imports the projection asonboardProjection, andsrc/init-command-router.cts:75wires theonboard:route that consumes it — confirming the Init Command Module integration (the ADR's literal handler name "initOnboard" is not itself a grep-matched symbol; the consumption is the router entry plus the destructured import).msd-core/workflows/onboard.md,commands/msd/onboard.md, andskills/msd-onboard/SKILL.mdall exist on disk, matching the "What stays OUTSIDE this Module" boundary.tests/onboard-command.test.cjs(25,129 bytes) contains named tests covering the load-bearing gate order —routes planning artifacts without PROJECT.md to partial planning,fast mode routes incomplete planning to partial-planning before the complete-map gate (regression #1990: fast map gate misroute)— vendor exclusion (ignores generated and vendor directories when detecting existing code), and package-manifest brownfield detection (treats package manifests as brownfield even without source files).- Six commits tagged
#1990landed the ADR, the projection module, and doc/index updates:3c7d722ed,e8fb05e96,d0b8eacd3,1171499f3,bc751a64e,192764f0c.
Governance state: Owning issue #1990 — CLOSED, stateReason COMPLETED, closed 2026-07-07T04:23:56Z.
Context
MSD already ships strong individual primitives for adopting an existing codebase: /msd:map-codebase (parallel codebase analysis), /msd:ingest-docs (classify and consolidate existing ADR/PRD/SPEC/RFC docs), and /msd:new-project (planning initialization). What it lacked was a single guided entry point that inspects a brownfield repository and tells the user which primitive runs first.
Left to prose alone, that ordering is ambiguous and unsafe: a user can initialize planning before a codebase map exists, skip relevant design docs, or overwrite/duplicate .planning/ context instead of reusing it. The ordering is not a matter of taste — it is a dependency graph (a map should exist before planning; existing design docs should be ingested before a fresh /msd:new-project; nothing should clobber an in-progress .planning/). A dependency graph that decides the next safe action from filesystem state is a projection, not something a workflow's natural-language instructions can evaluate reliably or test.
MSD already has the seam for this. The Init Command Module (src/init.cts → msd-core/bin/lib/init.cjs) owns the init.* family of query handlers that compose atomic queries into the flat JSON bundles that init workflows consume, alongside the projection-module precedent set by the Planning Path Projection Module (ADR-0006) and the Shell Command Projection Module (ADR-0009). Adding /msd:onboard as free-form workflow prose that scans the tree inline would put untested, non-deterministic filesystem logic in markdown — precisely the anti-pattern those projection modules exist to prevent.
Decision
Introduce the Existing Code Onboarding Module (implemented as the src/onboard-projection.cts → msd-core/bin/lib/onboard-projection.cjs projection) as the Seam that owns deterministic detection of brownfield repository state and the selection of the next onboarding action. It is a pure, side-effect-free projection consumed by the Init Command Module's initOnboard handler and rendered by the /msd:onboard workflow. It never writes; detection and route selection are a function of repository state only.
Detected state (inputs):
| Signal | Rule / invariant |
|---|---|
| Brownfield code present | Depth-capped recursive scan for source files (hasCodeFilesInternal) OR a recognized package manifest (hasPackageFileInternal). |
| Generated / vendor exclusion | Scan skips CODE_SCAN_SKIP_DIRS (node_modules, dist, build, .next, .nuxt, .svelte-kit, coverage, vendor, .venv, venv) so vendored trees never produce a false brownfield positive. |
| Codebase-map completeness | Whether .planning/codebase/ holds the canonical map artifacts. |
| Existing design docs | Presence of ADR/PRD/SPEC/RFC-style candidates (root, nested, and segment-based). |
| Partial planning state | Whether some but not all of PROJECT.md / REQUIREMENTS.md / ROADMAP.md / STATE.md exist. |
Route selection (output), ordered by dependency, not by convenience:
- Brownfield code without a complete
.planning/codebase/map → hand off to/msd:map-codebase(or/msd:map-codebase --fastin fast mode). - Design-doc candidates present and no project yet → offer
/msd:ingest-docsbefore/msd:new-project. - Otherwise →
/msd:new-project.
The gate order is load-bearing: partial-planning and fast-map-completeness are evaluated before the docs-ingest branch, so a half-mapped or half-initialized repo is never routed past the step it still owes. Handoff commands are runtime-formatted (buildHandoffCommands / formatMsdSlash) so the projected next command is correct for the installed runtime's slash syntax.
Safety invariants (the reason this is a Module, not a helper):
- Idempotent / no silent overwrite. Onboarding never mutates existing tracked
.planning/artifacts; re-running leaves them byte-unchanged. SUMMARY.mdis a trailing artifact..planning/onboarding/SUMMARY.mdis written only after project setup exists, and only if absent.- "Complete" is a conjunction. Onboarding does not report complete until
PROJECT.md,REQUIREMENTS.md,ROADMAP.md, andSTATE.mdall exist — no single-file short-circuit. - Text-mode parity.
--textrenders the same gate decisions as numbered plain-text prompts, so runtimes without an interactive picker get identical routing.
What stays OUTSIDE this Module
- The primitives themselves.
/msd:map-codebase,/msd:ingest-docs, and/msd:new-projectretain their own behavior; the Module only chooses and orders them. It projects the route; it does not re-implement the destinations. - Writing planning artifacts. All
.planning/writes remain owned by the destination commands and the Installer/planning modules. The projection is read-only. - The workflow's rendering.
msd-core/workflows/onboard.mdowns menu/gate presentation; the commandcommands/msd/onboard.md(and its skill mirror) owns delegation. The Module owns only the state→route decision they consume.
Consequences
- Brownfield onboarding becomes a single, testable entry point with deterministic routing, rather than order-of-operations folklore in prose. The projection is unit-tested (
tests/onboard-command.test.cjs) for brownfield/greenfield detection, vendor-dir exclusion, gate ordering (partial-planning before docs-ingest), idempotency/no-mutation, and runtime-formatted handoffs. - The Init Command Module gains one more heavyweight handler (
initOnboard) with the same{ data: <flat JSON> }contract as its siblings — no new dispatch shape. - New maintenance coupling, now explicit. The Module's completeness checks must track the canonical
.planning/codebase/artifact list and the routing targets' identities; if/msd:map-codebase//msd:ingest-docs//msd:new-projectchange their entry contracts, this projection must follow. This ADR records that coupling as the known cost of centralizing the routing decision (the alternative — duplicating the decision across each primitive — is worse). - No new runtime dependencies; no change to existing command semantics (additive).
Open questions
- Should codebase-map completeness be sourced from a single shared predicate (owned by the map module) rather than re-encoded here, so the two cannot drift?
- The
/msd:onboardworkflow sources itsmsd_runbootstrap from a sharedreferences/msd-run-resolver.mdsnippet rather than inlining it. If that delegation pattern is adopted by other workflows, it likely deserves its own short ADR — noting it here so the precedent is visible rather than silently established.
References
- ADR-0006 — Planning Path Projection Module (projection-module precedent for
.planningpath resolution). - ADR-0009 — Shell Command Projection Module (runtime-aware projection precedent).
- Init Command Module (
src/init.cts→msd-core/bin/lib/init.cjs) — owner of theinit.*handler family that consumes this projection viainitOnboard. CONTEXT.md§ Init Command Module — where the onboarding projection is registered.- Issue #1990 — feature spec and acceptance criteria.