refactor(project): merge DeployedProjectResource into ResolvedDeployedResource

[jariy17]

Summary

DeployedProjectResource and ResolvedDeployedResource share the same responsibility which is described the state of the deployed resource. Therefore, I have merged them. No behavior change should occur.

Verification

• tsc --noEmit — 0 errors
• bun test — 2598 pass / 0 fail (188 files)
• oxlint — clean

[agentcore-devx-automation]

Claude Security Review: no high-confidence findings. (run)

[agentcore-devx-automation]

AgentCore Harness Review

Verdict: Looks good

This is a clean, purely refactor-level type change:

• Merges the old DeployedProjectResource ({ resourceType, name, id }) and the previous ResolvedDeployedResource ({ id, target }) into a single self-describing ResolvedDeployedResource = { resourceType, name, id, target }.
• Updates the CDK backend to attach target to each resource, the manager to return the resource as-is from resolveDeployedResource, and adjusts tests to match.

I verified that:

• No references to DeployedProjectResource remain (grep is empty).
• Both consumers of resolveDeployedResource (runtime.tsx, harness.tsx) still access deployed.id and deployed.target.region, which the new shape supplies.
• screen.tsx still uses ResolvedDeployedResources (the envelope), which continues to expose deployed.target.region and deployed.resources[].id.
• The envelope's target and each resource's target are guaranteed to agree because both come from the same target value the manager passes into the backend; the doc comments call out the deliberate duplication.
• No new I/O, features, or user-visible behavior — no telemetry gap introduced.
• Tests were adjusted to the new type without additional mocking.

Nothing blocking; safe to merge.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 97.25%. Comparing base (27e3250) to head (e1f44c2).
⚠️ Report is 4 commits behind head on refactor.

Additional details and impacted files

@@            Coverage Diff            @@
##           refactor    #2160   +/-   ##
=========================================
  Coverage     97.25%   97.25%           
=========================================
  Files           508      508           
  Lines         33902    33902           
=========================================
  Hits          32972    32972           
  Misses          930      930           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[jariy17]

Bug bash: one real defect found and fixed (2998f2f8)

Two independent planners — a devil's-advocate pass and a Munger-inversion pass — were run over this diff. They converged on the same single defect (filed as INV-1 / B1 respectively). Everything else they raised was either already correct, inert, or a note for later.

The defect: target authority silently moved from the manager to the backend

The merge changed resolveDeployedResource from

return { id: resource.id, target: resolved.target };   // manager-authoritative

to return resource; — which hands back whatever target the backend stamped on the item. That is not a no-op: resolveExistingTarget resolves the target from aws-targets.json, and both invoke handlers pin the AWS region off this value (invoke/runtime.tsx, invoke/harness.tsx: ctx.withValue(RegionKey, deployed.target.region)). A backend echoing a stale or wrong target could therefore redirect a customer's invoke to the wrong region — impossible before this PR, because the manager's own resolved target always won.

Latent rather than live today (CdkBackend stamps the same object it was handed), but the merge created the degree of freedom, and neither the compiler nor the suite checked it.

Fix — the declared target overrides the item's copy:

if (resource) return { ...resource, target: resolved.target };

Doc comments deleted rather than reworded

Three comments misstated ownership and have been removed:

• the two blocks above ResolvedDeployedResource / ResolvedDeployedResources — one claimed a caller holding a single item needs no envelope, but the envelope still exists and invoke/screen.tsx is what actually reads it; the other restated what the types already show;
• the cdk.ts comment claiming "The backend already resolved target" — the manager resolved it and passed it in.

New test coverage

manager.test.ts had no describe for either method — the changed line was covered only transitively. Added FsProjectManager.resolveDeployedResource(s) with five cases:

Case	Asserts
divergence guard	backend claims us-east-1 while aws-targets.json declares eu-west-1 → singular returns eu-west-1
identity	a runtime and a harness sharing one name resolve to their own ids; resourceType/name echo the request
empty stack	zero resources still yields { resources: [], target }
throw path	ProjectStateError matching /is not deployed to target 'default'.*project deploy --target default/
invariant	every item's target is the same reference the envelope reports

The divergence guard was verified in both directions: it FAILS on 42138bcb and PASSES with the fix.

Mutation checks (each reverted)

#	Mutation	Result
M1	manager.tsx → return resolved.resources[0]!	🔴 RED (2 fail)
M2	cdk.ts → drop target from the item	🔴 RED — tsc TS2322 and tests; proves target is structurally required
M3	cdk.ts → stamp region: "eu-central-1"	🔴 RED (cdk.test.ts)
M4	invoke/index.test.tsx stub → drop target: input.target	🔴 tsc TS2322

M3 is worth a note: my first attempt mutated to us-east-1, which is exactly cdk.test.ts's TARGET.region, so the mutant object was identical to the original and the suite stayed green. That was an invalid mutation, not a coverage gap — re-running with eu-central-1 turns it red. Flagging it because a fixture whose value collides with a mutant is an easy way to fool yourself into trusting a green run.

Gates

tsc --noEmit clean · bun test 2603 pass / 0 fail · oxlint clean · prettier --check clean.

Still outstanding

Live-AWS (agentcore project deploy) and TUI-picker evidence are being produced separately against this commit and will be posted in a follow-up comment. Staying in draft until then.

[agentcore-devx-automation]

Claude Security Review: no high-confidence findings. (run)

[Hweinstock]

thanks for cleaning this up!

[agentcore-devx-automation]

Claude Security Review: no high-confidence findings. (run)