feat(ops): add portable War Room preflight command

[shivamsingh-007]

Summary

Add a portable
pm run ops:preflight\ command that checks War Room operational readiness before launching a mission.

Related Issue

Fixes #35

What changed

• New \scripts/ops-preflight.mjs\ — checks API health, mission status, pending approvals, local-agent/provider config, arsenal tool availability
• Added \ops:preflight\ npm script to \package.json\

Why this approach

Follows the same pattern as \scripts/doctor.mjs\ — fetch-based checks against the local War Room API with clean pass/fail/warn output. Supports --json\ for machine-readable output and --strict\ to fail on warnings.

Testing done

• 
  pm run typecheck\ passes
• Pre-existing test suite unaffected (no test changes needed)

Checklist

• No unrelated changes bundled in
• Linked issue referenced
• Commits signed off (DCO)

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has reached their quota limit.

[jmagly]

The preflight script looks good and is appropriately read-only (execFile, no shell, only hits the local API). One nit: the PR carries package-lock.json churn — 11 @emnapi/* entries flipped from dev/optional to peer — which looks like an npm-version lockfile rewrite rather than an intended change. Worth reverting the lockfile so the diff is just the script + the package.json ops:preflight line.

[shivamsingh-007]

Thanks @jmagly — good catch on the lockfile noise. Reverted package-lock.json to upstream main in b3bb807 so the diff is just the preflight script + package.json.

[jmagly]

Thanks for the quick lockfile revert (b3bb807) — the diff is clean now (just the script + the package.json line).

I went back and reviewed the script itself more thoroughly against the live server (started it and curled each endpoint the preflight hits). The good news: execFile/no-shell/local-only is solid, and the health and arsenal checks read the right shapes. But three checks read the wrong response shape and so silently pass regardless of actual state — which is the one thing a preflight can't do. Live responses vs what the script reads:

1. Mission-active check (line ~95). GET /api/mission/status returns {"active":false} (and {active:true, mission:{id,currentPhase}} when running). The script reads mission.data?.phase — that field doesn't exist, so missionActive is always falsy and "No active mission in progress" always passes, even mid-mission. This is the core check from #35, so it matters most.

const missionActive = mission.ok && mission.data?.active === true;
// (and mission.data?.mission?.id / .currentPhase for the detail string)

2. Pending-approvals check (line ~104). GET /api/approvals?status=pending returns {"approvals":[]}. The script does Array.isArray(approvals.data) ? approvals.data.length : approvals.data?.total || 0 — data is {approvals:[…]} (never an array) and there's no total, so pendingCount is always 0.

pendingCount = Array.isArray(approvals.data?.approvals) ? approvals.data.approvals.length : 0;

3. Local-agent check (line ~115). GET /api/agents/local/status returns {"connected":[…]} — an array. The script tests localAgent.data?.connected as a boolean, but [] is truthy in JS, so "Local agent connected" always passes even with zero agents (and data.name is undefined).

const agents = Array.isArray(localAgent.data?.connected) ? localAgent.data.connected : [];
const connected = agents.length > 0;   // name: agents[0]?.name

With those three fixed, this is a genuinely useful launch gate. Everything else (structure, exit codes, --json mode, the health/arsenal reads) looks right. Nice work.

[shivamsingh-007]

Thanks for the detailed review @jmagly. All three bugs are fixed in 07e7df8:

1. Mission check — now reads \mission.data?.active\ instead of \mission.data?.phase. The API returns {active: true/false}, not a phase field.

2. Pending approvals — now reads \�pprovals.data?.approvals.length\ instead of checking for a flat array or a \ otal\ field. The API returns {approvals: [...]}.

3. Local agent — now checks \connected.length > 0\ instead of relying on JS truthiness of an array (which is always truthy even when empty). Also reads the agent name from \connected[0]?.name.

Ready for another look.

[jmagly]

Please rebase this branch onto current main and remove the unrelated deletions before this can be reviewed.

The branch currently deletes benchmark corpora, benchmark scripts, provenance docs, and existing tests that are unrelated to the War Room preflight command. That is the primary blocker.

Suggested agent repair path:

1. Recreate or rebase the branch from current main.
2. Keep only the preflight command implementation, package script entry, and focused docs/tests for that command.
3. Restore all benchmark corpora/scripts, docs/VERIFIED_PROVENANCE.md, and unrelated tests.
4. Because this is advertised as portable, avoid POSIX-only command checks such as raw which; use Node-native checks or platform-specific fallbacks such as where.exe on Windows and command -v on POSIX.
5. Run git diff --name-status upstream/main...HEAD and confirm there are no unrelated D entries.