Files
kandev/docs/plans/go-dev-launcher/task-08-docs.md
T
2026-08-24 11:09:03 +00:00

2.1 KiB

id, title, status, wave, depends_on, plan, spec
id title status wave depends_on plan spec
08-docs Documentation and guidance updates done 3
06-makefile-cutover
plan.md ../../specs/platform/requirements/go-dev-launcher.md

Task 08: Documentation and guidance updates

Acceptance

  • No doc, skill, or agent-guidance file tells a reader that make dev runs through apps/cli/src/dev.ts or the TypeScript launcher.
  • The root CLAUDE.md "Tooling" section states that all four launch modes (dev, start, run, service) go through the Go binary, and that apps/cli is the npm publishing shim only.
  • docs/remote-cloud-environment.md reflects that make dev needs a Go toolchain plus pnpm-for-Vite, and no longer needs tsx.
  • docs/specs/INDEX.md lists the new spec.

Verification

grep -rn -E "apps/cli/src|cli/src/dev.ts|pnpm -C cli|tsx src/cli.ts" --include='*.md' --include='Makefile' --include='*.yml' . | grep -v docs/plans/

Returns nothing outside historical plan records under docs/plans/, which are implementation records and are intentionally not rewritten.

Files

  • CLAUDE.md (root) — "Tooling" section
  • docs/remote-cloud-environment.md
  • docs/specs/INDEX.md
  • docs/public/** — audit for launcher/dev-setup references via /docs-maintainer
  • Any .agents/skills/** or AGENTS.md hit from the grep above

Inputs

  • CLAUDE.md "Tooling" and "Maintaining This File" sections — the latter requires this update in the same PR as the code change.
  • apps/backend/CLAUDE.md — check whether the launcher package warrants a mention now that it owns every entrypoint.

Risks

  • docs/plans/{diagnostic-logging,port-collision-safety,feature-toggles-restart-action}/ reference apps/cli/src/* paths that will no longer exist. These are records of completed work, not live guidance — leave them alone and say so in the PR rather than rewriting history.
  • Consider whether "the Go launcher owns every entrypoint; apps/cli is a publishing shim" is a durable enough boundary to warrant an ADR via /record. It arguably is, since it constrains where future launcher features may be implemented.