feat(compat): salvage JSON-string submit payloads before failing schema validation (#185)

[OGtwelve]

Closes #185. XS P1 batch from the parity verification (#196).

Problem

Weak models frequently call the terminal submit tool with a JSON string (often markdown-fenced) instead of the object the schema wants. Live evidence (qwen/qwen3-coder:free): planner + worker + verify gate passed, two commits landed — then PR composition failed schema validation 3× (expected object, received string) and the group blocked at pr-open, one step from success.

Fix

submittedOutput, on a failed validation of a string payload, strips a single ```-fence and JSON.parses it, then re-validates: success is accepted as if the object had been submitted directly. Runs upstream of the #101 retry kernel, so retries only spend on genuinely wrong shapes. A non-string, non-parsing, or still-invalid value keeps the existing typed `invalid` result (with the parsed value's issues when it parsed but didn't validate). Zero cost to well-behaved models; provider-agnostic (no model sniffing).

Tests

• stringified-valid + fenced-valid salvage to ok; stringified-but-schema-invalid stays invalid with the parsed-object issues; non-JSON string and object paths unchanged.
• Green: typecheck (both) + test:node (compat 258) + biome.

Part of #196.

Summary by CodeRabbit

• Bug Fixes
  • Improved submitted output handling when tool input is provided as a JSON string.
  • Added support for JSON wrapped in fenced code blocks.
  • Valid JSON strings are now correctly parsed and validated against the expected schema.
  • Invalid parsed values now report validation issues against the parsed content.

[OGtwelve]

@coderabbitai full review

[coderabbitai]

✅ Action performed

Full review finished.

---

You're currently rate limited under our Fair Usage Limits Policy. Your recent PR review activity is in the 95th percentile or higher among CodeRabbit users, so adaptive limits apply. Your next review will be available in 46 minutes.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yml

Review profile: CHILL

Plan: Pro

Run ID: 9f789c12-7a8c-4986-92bd-64377142072c

📥 Commits

Reviewing files that changed from the base of the PR and between 277c344 and a15e36d.

📒 Files selected for processing (2)

• packages/ai-claude-compat/src/subagent.test.ts
• packages/ai-claude-compat/src/subagent.ts

---

📝 Walkthrough

Walkthrough

submittedOutput now salvages stringified or fenced JSON submit payloads, revalidates parsed values against the schema, and reports parsed validation issues when necessary. Tests cover valid and invalid cases.

Changes

JSON submit payload salvage

Layer / File(s)	Summary
Salvage and revalidate submit inputs packages/ai-claude-compat/src/subagent.ts	submittedOutput parses string inputs, tolerates a single fenced JSON wrapper, and revalidates parsed values while preserving existing behavior for non-string or unparsable inputs.
Validate recovered payload behavior packages/ai-claude-compat/src/subagent.test.ts	Tests cover valid stringified JSON, fenced JSON, and parsed payloads that fail schema validation.

Estimated code review effort: 2 (Simple) | ~10 minutes

Sequence Diagram(s)

sequenceDiagram
  participant SubmitToolCall
  participant submittedOutput
  participant salvageJsonString
  participant outputSchema
  SubmitToolCall->>submittedOutput: Provide call.input
  submittedOutput->>outputSchema: Validate original input
  outputSchema-->>submittedOutput: Return validation failure
  submittedOutput->>salvageJsonString: Parse string or fenced JSON
  salvageJsonString-->>submittedOutput: Return parsed value
  submittedOutput->>outputSchema: Validate parsed value
  outputSchema-->>submittedOutput: Return typed result or parsed issues

Loading

Possibly related PRs

• developerz-ai/ai-task-master#133: Establishes the typed submittedOutput and schema-retry handling extended by this change.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title clearly matches the main change: salvaging JSON-string submit payloads before schema validation fails.
Linked Issues check	✅ Passed	The implementation and tests align with #185 by accepting valid JSON strings, fenced JSON, and preserving invalid behavior otherwise.
Out of Scope Changes check	✅ Passed	The changes stay focused on the linked compatibility fix and its tests, with no obvious unrelated additions.
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check.

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch fix/salvage-json-submit-185

---

_{Comment @coderabbitai help to get the list of available commands.}

[OGtwelve]

@coderabbitai full review

[coderabbitai]

✅ Action performed

Full review finished.