fix(workflows): lint unquoted string references in JSON fields and surface lint on deploy

[waleedlatif1]

Summary

• Lint now flags a block reference written unquoted in a JSON editor field (Table row data, API body, …) when the referenced output is declared string. Outside Function code references are substituted as raw text, so {"id": <start.order_id>} becomes invalid JSON and the block failed only at run time, while the edit result's lint, workflows lint, and deploy all reported clean. Numbers, booleans, objects and undeclared types are not flagged.
• The finding reuses unresolvedReferences (kind: 'block-output', reason unquoted-json-string: …), so it appears in the lint field of every graph write and in workflows lint with no wire-shape change. Tokens come from the runtime's own reference scanner; only tokens outside JSON string literals (escapes handled) are considered.
• POST /api/v2/workflows/{id}/deploy lints the exact version it just made live and, if anything is found, appends one entry to the existing warnings array. Deploy never refuses or fails on lint: findings stay advisory, and a lint failure is logged and ignored.
• unresolvedReferences description updated (OpenAPI regenerated); the lint summary header for block-output findings reads "will not work as written".

Type of Change

• Bug fix

Testing

• New json-literal-refs.test.ts (unquoted string refs flagged; quoted, escaped-quote, JSON-typed, unknown-type, non-JSON-field and non-string values not flagged), a lint-report wiring test, and deploy use-case tests (warning added as the deploying user, clean graph adds nothing, lint failure never fails the deploy). Each failed before the fix; escape handling and the deploy try/catch were each reverted and watched go red.
• bunx vitest run lib/workflows lib/mothership/agent-cli/engines app/api/v2/workflows (182 files, 2757 tests), bun run lint, bun run type-check, bun run check:audits, bun run docs-manifest:check.
• Sim CLI command inventory is byte-identical to staging (no CLI contract change).

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing (new tests pass the test-audit authoring gate)
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment

Project	Deployment	Actions	Updated
docs	Skipped		Oct 8, 2026 4:01pm UTC

[cubic-dev-ai]

No issues found across 9 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

_{Turn on auto-fix | Re-trigger cubic}

[greptile-apps]

Confidence Score: 5/5

[Medium risk] Adds lint detection for unquoted string references in JSON fields.

The PR appears safe to merge; the canonical filter fix introduces no actionable issue.

What we checked:

• Inactive filters cause warnings: The basic builder stores an array. The lint check skips non-string values, so it does not scan those rules as JSON text.

Summary

The PR warns about unquoted string references in JSON fields and adds advisory lint findings to deploy responses.

• The latest change reads JSON editors through canonicalParamId, restoring warnings for Table advanced filters.
• Deploy warnings now describe the version being published, including pending activation.
• All four earlier, unnumbered findings are addressed: unused fields are skipped, lint uses the acting subject, the subject test checks returned behavior, and advanced filters use their sent parameter.
• No new actionable issues or repository-rule violations were found.

_{Reviews (3) · Last reviewed commit: "fix(workflows): check JSON editors sent ..." · Reviewed by Greptile}

[waleedlatif1]

@greptile

[waleedlatif1]

@cubic-dev-ai review this PR

[cubic-dev-ai]

@cubic-dev-ai review this PR

@waleedlatif1 I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

All reported issues were addressed across 9 files

<sub>Reply with feedback, questions, or to request a fix.

Turn on auto-fix | Re-trigger cubic</sub>

[waleedlatif1]

@greptile

[waleedlatif1]

@cubic-dev-ai review this PR

[cubic-dev-ai]

@cubic-dev-ai review this PR

@waleedlatif1 I have started the AI code review. It will take a few minutes to complete.

[cubic-dev-ai]

No issues found across 9 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

_{Turn on auto-fix | Re-trigger cubic}