Plan - Prisma — ContextQMD

Dispatch plan

Outcome: The default Prisma CLI resolves its existing identity unchanged, while a buildable prisma7 binary selects the private one-shot identity and delegates all arguments, completion dispatch, exit behavior, and subprocess re-entry to its declared Prisma CLI dependency.
Builds on: The slice spec's chosen identity-transport and dependency-wrapper design.
Hands to: A runnable unpublished wrapper with a frozen identity contract, no copied CLI implementation, and focused tests proving default Prisma and wrapped Prisma select the intended identities.
Focus: Identity resolution lifecycle, marker consumption, executable delegation, package skeleton/build, argument and completion forwarding, and default-behavior regression coverage. Public root/config parity and packed-manifest proof belong to dispatch 2.

Outcome: prisma7, prisma7/config, and prisma7/package.json expose the intended forwarded runtime/type contracts, and a packed artifact records the exact matching Prisma dependency and resolves it correctly beside a different direct root Prisma version.
Builds on: Dispatch 1's runnable wrapper and frozen identity contract.
Hands to: A complete unpublished slice artifact satisfying the wrapper package contract, ready for exhaustive identity propagation in the next project slice.
Focus: Export map and files manifest, root/config runtime and type forwarding, exact workspace-to-packed dependency semantics, package inspection, and local side-by-side resolution tests. Exhaustive CLI wording and all publication automation remain out of scope.

Outcome: CLI identity is derived solely from the normalized stem of process.argv[1]; the distinctive prisma7.js target selects prisma7 across supported package-manager launch forms, while ordinary Prisma and unsupported caller-defined entrypoints default to prisma.
Builds on: Dispatch 2's complete wrapper package and the cross-package-manager /tmp probe that falsified the marker/global-symbol design.
Hands to: A simpler reviewer-verified wrapper with no identity environment marker, cleanup, global symbol, package metadata lookup, or cross-bundle mutable state, ready for the next project slice.
Focus: Replace marker/global identity transport, give the wrapper a distinctive executable target, update emitted/forwarding/packed contract tests, and regression-pin normalized shim-vs-target names. Exhaustive user-facing identity propagation remains the next slice.

Outcome: The prisma7 executable directly loads prisma/build/index.js; the production-only delegation helper and callback-injection tests are removed, while built/packed executable coverage proves real dependency-edge loading, argv preservation, and exit propagation.
Builds on: Dispatch 3's distinctive prisma7.js target and accepted packed package contract.
Hands to: A minimal wrapper whose production code contains no abstraction beyond the required dependency load, backed by behavior-level executable evidence.
Focus: Inline the literal Prisma entrypoint load, delete the helper and tautological unit seam, and replace its claims with a network-free built/packed wrapper execution fixture. No broader branding or publication work.

Outcome: All unit/contract tests introduced for the wrapper and identity seam are replaced by one network-free end-to-end project test that typechecks and loads prisma7/config, proves the CLI uses that config, runs prisma7 --version, runs prisma7 generate, and executes the generated client successfully.
Builds on: Dispatch 4's minimal direct wrapper and the real package/config/generator implementations already present in the built workspace.
Hands to: A review surface with one user-visible behavioral proof instead of implementation-specific package, dispatcher, identity, and forwarding assertions.
Focus: Delete package-contract.test.ts, cli-distribution-identity.vitest.ts, and bin-dispatcher.vitest.ts; keep only setup and assertions required for config import/typechecking, version execution, generation through the chosen config, and a working generated client.

Outcome: The single compatibility scenario is an actual auto-discovered Docker E2E under packages/client/tests/e2e, installed from the packed prisma7 and workspace dependency tarballs exactly like shipped client fixtures; packages/prisma7 contains no Vitest test or package-level test job.
Builds on: Dispatch 5's settled three-outcome scenario and the existing client E2E runner's package, engine, TypeScript, and SQLite environment.
Hands to: A side-by-side wrapper proven through the repository's standard shipped-artifact E2E boundary rather than a package-local subprocess simulation.
Focus: Remove packages/prisma7/src/e2e.test.ts, its test-only dev dependencies/script, and its standalone workflow entries; add one prisma7-compatibility fixture with _steps.ts, config consumer, non-default schema/output, and SQLite client smoke; teach the tarball mount/rewrite logic that unscoped prisma7 packs as prisma7-0.0.0.tgz rather than the scoped-folder convention. No unit/contract tests.

Outcome: The distribution identity is the minimal 'prisma' | 'prisma7' value downstream consumers actually need, the E2E prepares its schema through prisma7 db push rather than raw SQL, and failed wrapper builds do not mask esbuild diagnostics with a chmod ENOENT.
Builds on: Dispatch 6's reviewed packed-artifact client E2E and final wrapper build shape.
Hands to: A reviewer-ready PR with all still-applicable human and bot inline feedback addressed or explicitly classified obsolete/rejected.
Focus: Remove the redundant identity object/map, add db push --force-reset to E2E setup and simplify the SQLite smoke, guard the executable plugin on build errors, rerun the focused Docker E2E/build gates, and leave rejected prepack plus comments on deleted tests untouched in code.