improvement(ship): pre-flight regenerate artifacts + parallel audit suite

.agents/skills/ship/SKILL.md

@@ -35,10 +35,47 @@ When the user runs `/ship`:
 5. **Run migration safety** — only if the diff touches `packages/db/migrations/**` or `packages/db/schema.ts`:
  - Run `/db-migrate` to review the migration for zero-downtime safety (expand/contract phasing, backward-compatibility with the deployed app version).
  - `bun run check:migrations origin/staging` must pass (staging is the PR base). Do not silence a flagged statement with a `-- migration-safe:` annotation unless `/db-migrate` confirmed the old code no longer depends on it; otherwise split the destructive change into a later deploy.
-6. **Run pre-ship checks** from the repo root before staging:
- - `bun run lint` to fix formatting issues
- - `bun run check:api-validation:strict` to catch boundary contract failures before CI
-7. **Stage and commit** the changes with the generated message
+6. **Run pre-ship checks** from the repo root before staging. This has two phases: first **regenerate** every committed artifact so generated files never drift into a CI failure (this is what catches things like `agent-stream-docs` going stale after a `models.ts` edit), then run the **full audit suite** CI's `Lint and Test` job enforces. Both phases parallelize — but only across commands that write **disjoint** outputs — and a bare `wait` swallows child exit codes, so both phases below explicitly collect each job's status and abort ship if any failed.
+
+ **Phase A — regenerate the always-in-repo committed artifacts (parallel), then let step 7 stage whatever changed.** Regenerate only the generators whose inputs live entirely in this repo and that any ordinary code change can drift — `agent-stream-docs:generate` (derives from the provider model registry) and `skills:sync` (derives from `.agents/skills/**`). They write disjoint trees (`apps/docs/…/agent.mdx` vs the `.claude`/`.cursor` command projections), so they parallelize safely, and each is idempotent (a no-op when already in sync):
+ ```bash
+ rm -f /tmp/ship-gen-results
+ for g in agent-stream-docs:generate skills:sync; do
+ ( bun run "$g" >"/tmp/ship-gen-${g//:/-}.log" 2>&1; echo "$? $g" >>/tmp/ship-gen-results ) &
+ done
+ wait
+ # any non-zero line is a FAILED generator — read /tmp/ship-gen-<name>.log and fix before shipping;
+ # a silently-failed generate leaves a stale artifact that Phase B / CI then rejects.
+ # The `exit 1` makes this block itself exit non-zero on failure, so anything gating on the
+ # command's status (an agent, or a wrapping script) actually stops — do NOT collapse it to
+ # `grep … && echo ❌ || echo ✅`, which always exits 0 and silently lets ship continue.
+ if grep -vE '^0 ' /tmp/ship-gen-results; then echo "❌ generator(s) failed — do not ship"; exit 1; fi
+ echo "✅ artifacts regenerated"
+ ```
+ Then `git status --short` to see what regenerated — those files must be staged in step 7 alongside your own changes.
+
+ **Do NOT blanket-run the domain generators here.** `mship:generate` (`generate-mship-contracts.ts`) is an **umbrella** that drives all nine mothership contract generators (`mship-contracts`, `billing-protocol-contract`, `mship-tools`, the four `trace-*`, `metrics-contract`, `vfs-snapshot-contract`) and biome-formats `apps/sim/lib/copilot/generated/` — never run it *and* its constituents (they write the same files and corrupt each other in parallel), and never run it on an ordinary ship: it reads an **external** copilot-contract source that isn't checked out in most worktrees, so it hard-fails with `ENOENT` and would abort ship for an unrelated reason. `generate:pi-model-catalog` (under `apps/sim`) likewise regenerates from the installed Pi package, not repo source. Only when **this PR's diff actually touches** a domain generator's input do you regenerate it deliberately and run its matching `:check` (`bun run mship:check` / the individual `*:check`) — with the external source present.
+
+ **Phase B — run lint + every audit CI enforces, in parallel, and abort ship if any fails.** `bun run lint` first (it autofixes formatting and mutates files, so don't parallelize it with the read-only audits), then fan the rest out and collect exit codes. This is exactly the read-only audit set from CI's `Lint and Test` job (all in-repo, runnable in any worktree):
+ ```bash
+ # autofix formatting first (mutating; not parallel-safe with the audits). Gate its exit too —
+ # a non-zero lint (unfixable errors) must abort before the audits run, not be ignored.
+ bun run lint || { echo "❌ lint failed — do not ship"; exit 1; }
+ rm -f /tmp/ship-audit-results
+ for s in check:boundaries check:api-validation:strict check:utils check:zustand-v5 \
+ check:react-query check:client-boundary check:bare-icons check:icon-paths \
+ check:realtime-prune skills:check agent-stream-docs:check; do
+ ( bun run "$s" >"/tmp/ship-audit-${s//:/-}.log" 2>&1; echo "$? $s" >>/tmp/ship-audit-results ) &
+ done
+ wait
+ # any non-zero line is a failing audit — read its /tmp/ship-audit-<name>.log and fix before shipping.
+ # `exit 1` on failure preserves the original sequential checks' semantics (their non-zero exit is
+ # what an agent gates on); never use `grep … && echo ❌ || echo ✅` here — it always exits 0.
+ if grep -vE '^0 ' /tmp/ship-audit-results; then echo "❌ audit(s) failed — do not ship"; exit 1; fi
+ echo "✅ all audits passed"
+ ```
+ If Phase A regenerated a file, its matching `:check` in Phase B now passes trivially — that parity is the point. Do not ship with any generator or audit failing; fix the cause (never silence it) and re-run. `check:migrations` and `type-check` are covered by steps 5 and CI respectively and are not repeated here.
+7. **Stage and commit** the changes with the generated message — including any files Phase A regenerated in step 6
 8. **Push to origin** using the current branch name — `--force-with-lease` if step 2's sync
  check did any history rewrite (a clean rebase or a cherry-pick rebuild) on a branch that had
  already been pushed once; a plain push would be rejected in exactly the polluted-remote case

.claude/commands/ship.md

@@ -34,10 +34,47 @@ When the user runs `/ship`:
 5. **Run migration safety** — only if the diff touches `packages/db/migrations/**` or `packages/db/schema.ts`:
  - Run `/db-migrate` to review the migration for zero-downtime safety (expand/contract phasing, backward-compatibility with the deployed app version).
  - `bun run check:migrations origin/staging` must pass (staging is the PR base). Do not silence a flagged statement with a `-- migration-safe:` annotation unless `/db-migrate` confirmed the old code no longer depends on it; otherwise split the destructive change into a later deploy.
-6. **Run pre-ship checks** from the repo root before staging:
- - `bun run lint` to fix formatting issues
- - `bun run check:api-validation:strict` to catch boundary contract failures before CI
-7. **Stage and commit** the changes with the generated message
+6. **Run pre-ship checks** from the repo root before staging. This has two phases: first **regenerate** every committed artifact so generated files never drift into a CI failure (this is what catches things like `agent-stream-docs` going stale after a `models.ts` edit), then run the **full audit suite** CI's `Lint and Test` job enforces. Both phases parallelize — but only across commands that write **disjoint** outputs — and a bare `wait` swallows child exit codes, so both phases below explicitly collect each job's status and abort ship if any failed.
+
+ **Phase A — regenerate the always-in-repo committed artifacts (parallel), then let step 7 stage whatever changed.** Regenerate only the generators whose inputs live entirely in this repo and that any ordinary code change can drift — `agent-stream-docs:generate` (derives from the provider model registry) and `skills:sync` (derives from `.agents/skills/**`). They write disjoint trees (`apps/docs/…/agent.mdx` vs the `.claude`/`.cursor` command projections), so they parallelize safely, and each is idempotent (a no-op when already in sync):
+ ```bash
+ rm -f /tmp/ship-gen-results
+ for g in agent-stream-docs:generate skills:sync; do
+ ( bun run "$g" >"/tmp/ship-gen-${g//:/-}.log" 2>&1; echo "$? $g" >>/tmp/ship-gen-results ) &
+ done
+ wait
+ # any non-zero line is a FAILED generator — read /tmp/ship-gen-<name>.log and fix before shipping;
+ # a silently-failed generate leaves a stale artifact that Phase B / CI then rejects.
+ # The `exit 1` makes this block itself exit non-zero on failure, so anything gating on the
+ # command's status (an agent, or a wrapping script) actually stops — do NOT collapse it to
+ # `grep … && echo ❌ || echo ✅`, which always exits 0 and silently lets ship continue.
+ if grep -vE '^0 ' /tmp/ship-gen-results; then echo "❌ generator(s) failed — do not ship"; exit 1; fi
+ echo "✅ artifacts regenerated"
+ ```
+ Then `git status --short` to see what regenerated — those files must be staged in step 7 alongside your own changes.
+
+ **Do NOT blanket-run the domain generators here.** `mship:generate` (`generate-mship-contracts.ts`) is an **umbrella** that drives all nine mothership contract generators (`mship-contracts`, `billing-protocol-contract`, `mship-tools`, the four `trace-*`, `metrics-contract`, `vfs-snapshot-contract`) and biome-formats `apps/sim/lib/copilot/generated/` — never run it *and* its constituents (they write the same files and corrupt each other in parallel), and never run it on an ordinary ship: it reads an **external** copilot-contract source that isn't checked out in most worktrees, so it hard-fails with `ENOENT` and would abort ship for an unrelated reason. `generate:pi-model-catalog` (under `apps/sim`) likewise regenerates from the installed Pi package, not repo source. Only when **this PR's diff actually touches** a domain generator's input do you regenerate it deliberately and run its matching `:check` (`bun run mship:check` / the individual `*:check`) — with the external source present.
+
+ **Phase B — run lint + every audit CI enforces, in parallel, and abort ship if any fails.** `bun run lint` first (it autofixes formatting and mutates files, so don't parallelize it with the read-only audits), then fan the rest out and collect exit codes. This is exactly the read-only audit set from CI's `Lint and Test` job (all in-repo, runnable in any worktree):
+ ```bash
+ # autofix formatting first (mutating; not parallel-safe with the audits). Gate its exit too —
+ # a non-zero lint (unfixable errors) must abort before the audits run, not be ignored.
+ bun run lint || { echo "❌ lint failed — do not ship"; exit 1; }
+ rm -f /tmp/ship-audit-results
+ for s in check:boundaries check:api-validation:strict check:utils check:zustand-v5 \
+ check:react-query check:client-boundary check:bare-icons check:icon-paths \
+ check:realtime-prune skills:check agent-stream-docs:check; do
+ ( bun run "$s" >"/tmp/ship-audit-${s//:/-}.log" 2>&1; echo "$? $s" >>/tmp/ship-audit-results ) &
+ done
+ wait
+ # any non-zero line is a failing audit — read its /tmp/ship-audit-<name>.log and fix before shipping.
+ # `exit 1` on failure preserves the original sequential checks' semantics (their non-zero exit is
+ # what an agent gates on); never use `grep … && echo ❌ || echo ✅` here — it always exits 0.
+ if grep -vE '^0 ' /tmp/ship-audit-results; then echo "❌ audit(s) failed — do not ship"; exit 1; fi
+ echo "✅ all audits passed"
+ ```
+ If Phase A regenerated a file, its matching `:check` in Phase B now passes trivially — that parity is the point. Do not ship with any generator or audit failing; fix the cause (never silence it) and re-run. `check:migrations` and `type-check` are covered by steps 5 and CI respectively and are not repeated here.
+7. **Stage and commit** the changes with the generated message — including any files Phase A regenerated in step 6
 8. **Push to origin** using the current branch name — `--force-with-lease` if step 2's sync
  check did any history rewrite (a clean rebase or a cherry-pick rebuild) on a branch that had
  already been pushed once; a plain push would be rejected in exactly the polluted-remote case

.cursor/commands/ship.md

@@ -29,10 +29,47 @@ When the user runs `/ship`:
 5. **Run migration safety** — only if the diff touches `packages/db/migrations/**` or `packages/db/schema.ts`:
  - Run `/db-migrate` to review the migration for zero-downtime safety (expand/contract phasing, backward-compatibility with the deployed app version).
  - `bun run check:migrations origin/staging` must pass (staging is the PR base). Do not silence a flagged statement with a `-- migration-safe:` annotation unless `/db-migrate` confirmed the old code no longer depends on it; otherwise split the destructive change into a later deploy.
-6. **Run pre-ship checks** from the repo root before staging:
- - `bun run lint` to fix formatting issues
- - `bun run check:api-validation:strict` to catch boundary contract failures before CI
-7. **Stage and commit** the changes with the generated message
+6. **Run pre-ship checks** from the repo root before staging. This has two phases: first **regenerate** every committed artifact so generated files never drift into a CI failure (this is what catches things like `agent-stream-docs` going stale after a `models.ts` edit), then run the **full audit suite** CI's `Lint and Test` job enforces. Both phases parallelize — but only across commands that write **disjoint** outputs — and a bare `wait` swallows child exit codes, so both phases below explicitly collect each job's status and abort ship if any failed.
+
+ **Phase A — regenerate the always-in-repo committed artifacts (parallel), then let step 7 stage whatever changed.** Regenerate only the generators whose inputs live entirely in this repo and that any ordinary code change can drift — `agent-stream-docs:generate` (derives from the provider model registry) and `skills:sync` (derives from `.agents/skills/**`). They write disjoint trees (`apps/docs/…/agent.mdx` vs the `.claude`/`.cursor` command projections), so they parallelize safely, and each is idempotent (a no-op when already in sync):
+ ```bash
+ rm -f /tmp/ship-gen-results
+ for g in agent-stream-docs:generate skills:sync; do
+ ( bun run "$g" >"/tmp/ship-gen-${g//:/-}.log" 2>&1; echo "$? $g" >>/tmp/ship-gen-results ) &
+ done
+ wait
+ # any non-zero line is a FAILED generator — read /tmp/ship-gen-<name>.log and fix before shipping;
+ # a silently-failed generate leaves a stale artifact that Phase B / CI then rejects.
+ # The `exit 1` makes this block itself exit non-zero on failure, so anything gating on the
+ # command's status (an agent, or a wrapping script) actually stops — do NOT collapse it to
+ # `grep … && echo ❌ || echo ✅`, which always exits 0 and silently lets ship continue.
+ if grep -vE '^0 ' /tmp/ship-gen-results; then echo "❌ generator(s) failed — do not ship"; exit 1; fi
+ echo "✅ artifacts regenerated"
+ ```
+ Then `git status --short` to see what regenerated — those files must be staged in step 7 alongside your own changes.
+
+ **Do NOT blanket-run the domain generators here.** `mship:generate` (`generate-mship-contracts.ts`) is an **umbrella** that drives all nine mothership contract generators (`mship-contracts`, `billing-protocol-contract`, `mship-tools`, the four `trace-*`, `metrics-contract`, `vfs-snapshot-contract`) and biome-formats `apps/sim/lib/copilot/generated/` — never run it *and* its constituents (they write the same files and corrupt each other in parallel), and never run it on an ordinary ship: it reads an **external** copilot-contract source that isn't checked out in most worktrees, so it hard-fails with `ENOENT` and would abort ship for an unrelated reason. `generate:pi-model-catalog` (under `apps/sim`) likewise regenerates from the installed Pi package, not repo source. Only when **this PR's diff actually touches** a domain generator's input do you regenerate it deliberately and run its matching `:check` (`bun run mship:check` / the individual `*:check`) — with the external source present.
+
+ **Phase B — run lint + every audit CI enforces, in parallel, and abort ship if any fails.** `bun run lint` first (it autofixes formatting and mutates files, so don't parallelize it with the read-only audits), then fan the rest out and collect exit codes. This is exactly the read-only audit set from CI's `Lint and Test` job (all in-repo, runnable in any worktree):
+ ```bash
+ # autofix formatting first (mutating; not parallel-safe with the audits). Gate its exit too —
+ # a non-zero lint (unfixable errors) must abort before the audits run, not be ignored.
+ bun run lint || { echo "❌ lint failed — do not ship"; exit 1; }
+ rm -f /tmp/ship-audit-results
+ for s in check:boundaries check:api-validation:strict check:utils check:zustand-v5 \
+ check:react-query check:client-boundary check:bare-icons check:icon-paths \
+ check:realtime-prune skills:check agent-stream-docs:check; do
+ ( bun run "$s" >"/tmp/ship-audit-${s//:/-}.log" 2>&1; echo "$? $s" >>/tmp/ship-audit-results ) &
+ done
+ wait
+ # any non-zero line is a failing audit — read its /tmp/ship-audit-<name>.log and fix before shipping.
+ # `exit 1` on failure preserves the original sequential checks' semantics (their non-zero exit is
+ # what an agent gates on); never use `grep … && echo ❌ || echo ✅` here — it always exits 0.
+ if grep -vE '^0 ' /tmp/ship-audit-results; then echo "❌ audit(s) failed — do not ship"; exit 1; fi
+ echo "✅ all audits passed"
+ ```
+ If Phase A regenerated a file, its matching `:check` in Phase B now passes trivially — that parity is the point. Do not ship with any generator or audit failing; fix the cause (never silence it) and re-run. `check:migrations` and `type-check` are covered by steps 5 and CI respectively and are not repeated here.
+7. **Stage and commit** the changes with the generated message — including any files Phase A regenerated in step 6
 8. **Push to origin** using the current branch name — `--force-with-lease` if step 2's sync
  check did any history rewrite (a clean rebase or a cherry-pick rebuild) on a branch that had
  already been pushed once; a plain push would be rejected in exactly the polluted-remote case