fix(code-mode): validate tool input/output against schemas

[anatolzak]

🎯 Changes

Tools used inside code mode never validated their inputSchema/outputSchema. The normal agent-loop path validates args before execute and the result after (parseWithStandardSchema in activities/chat/tools/tool-calls.ts), but ai-code-mode's toolToBinding only used the schema to build the prompt and forwarded the sandbox's raw args straight through:

const execute = (args, context) => Promise.resolve(toolExecute(args, context))
So external_* tools received unvalidated, un-coerced input and their output went unchecked — a provided schema was effectively just documentation.

This change runs the same Standard Schema validation in toolToBinding's execute: defaults/coercions now apply, and invalid input/output throws an agent-readable error (Input/Output validation failed for tool : …) that surfaces back into the sandbox so the model can self-correct. Added unit tests covering input rejection, coercion/defaults, and output validation.

closes#793

Note: code mode has no existing E2E harness (no driver/route/page in testing/e2e). I've added unit coverage but flagging since CONTRIBUTING lists E2E as required.

✅ Checklist

• I have followed the steps in the Contributing guide.
• I have tested this code locally with pnpm run test:pr.

🚀 Release Impact

• This change affects published code, and I have generated a changeset.
• This change is docs/CI/dev-only (no release).

Summary by CodeRabbit

• Bug Fixes
  • Code mode tools now validate inputs and outputs against their declared schemas to ensure data integrity.
  • Schema-based defaults and coercions are automatically applied to tool inputs for improved consistency.
  • Invalid inputs or outputs generate clear, actionable error messages to aid in debugging and error resolution.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Pro

Run ID: bca0ef7b-d609-471d-8639-dc32ae277f26

📥 Commits

Reviewing files that changed from the base of the PR and between 2cb0313 and 8873b33.

📒 Files selected for processing (3)

• .changeset/code-mode-tool-validation.md
• packages/ai-code-mode/src/bindings/tool-to-binding.ts
• packages/ai-code-mode/tests/tool-to-binding.test.ts

---

📝 Walkthrough

Walkthrough

toolToBinding in ai-code-mode now imports isStandardSchema and parseWithStandardSchema from @tanstack/ai and uses them inside execute to validate and coerce tool inputs before calling the underlying tool, and to validate outputs afterward. Invalid inputs or outputs throw agent-readable errors. Three new tests and a changeset entry accompany the change.

Changes

Code Mode Tool Schema Validation

Layer / File(s)	Summary
Standard Schema validation in toolToBinding execute packages/ai-code-mode/src/bindings/tool-to-binding.ts	Adds isStandardSchema and parseWithStandardSchema imports; rewrites execute as async to conditionally parse/coerce args through parseWithStandardSchema before calling the tool, and validates the output against outputSchema when present, rethrowing failures as "Input validation failed for tool <name>" or "Output validation failed for tool <name>".
Validation tests and Changeset entry packages/ai-code-mode/tests/tool-to-binding.test.ts, .changeset/code-mode-tool-validation.md	Adds three test cases covering input rejection on schema mismatch, Zod default application before execute is called, and output rejection when the returned shape does not conform; adds a patch Changeset entry documenting the fix.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Poem

🐇 A schema once ignored in code,
Now guards each input down the road.
Defaults applied, bad types refused,
No silent pass-through left confused.
Validation hops both in and out —
This rabbit fixed it, have no doubt! ✨

🚥 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 accurately summarizes the main change: adding validation for tool input/output against schemas in code mode, which is the core objective of the PR.
Linked Issues check	✅ Passed	The PR fully addresses issue #793 by implementing Standard Schema validation in toolToBinding's execute function, ensuring input coercion and validation occur before tool execution, and output validation occurs afterward.
Out of Scope Changes check	✅ Passed	All changes are directly scoped to fixing code-mode tool validation against schemas as described in the linked issue; no unrelated modifications are present.
Docstring Coverage	✅ Passed	Docstring coverage is 100.00% which is sufficient. The required threshold is 80.00%.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

🧪 Generate unit tests (beta)

• Create PR with unit tests

Warning

Tools execution failed with the following error:

Failed to run tools: 13 INTERNAL: Received RST_STREAM with code 2 (Internal server error)

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

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

[nx-cloud]

View your CI Pipeline Execution ↗ for commit 8873b33

Command	Status	Duration	Result
nx run-many --targets=build --exclude=examples/...	✅ Succeeded	13s	View ↗

---

☁️ Nx Cloud last updated this comment at 2026-06-19 13:54:53 UTC

[pkg-pr-new]

Open in StackBlitz

@tanstack/ai

npm i https://pkg.pr.new/@tanstack/ai@794

@tanstack/ai-angular

npm i https://pkg.pr.new/@tanstack/ai-angular@794

@tanstack/ai-anthropic

npm i https://pkg.pr.new/@tanstack/ai-anthropic@794

@tanstack/ai-client

npm i https://pkg.pr.new/@tanstack/ai-client@794

@tanstack/ai-code-mode

npm i https://pkg.pr.new/@tanstack/ai-code-mode@794

@tanstack/ai-code-mode-skills

npm i https://pkg.pr.new/@tanstack/ai-code-mode-skills@794

@tanstack/ai-devtools-core

npm i https://pkg.pr.new/@tanstack/ai-devtools-core@794

@tanstack/ai-elevenlabs

npm i https://pkg.pr.new/@tanstack/ai-elevenlabs@794

@tanstack/ai-event-client

npm i https://pkg.pr.new/@tanstack/ai-event-client@794

@tanstack/ai-fal

npm i https://pkg.pr.new/@tanstack/ai-fal@794

@tanstack/ai-gemini

npm i https://pkg.pr.new/@tanstack/ai-gemini@794

@tanstack/ai-grok

npm i https://pkg.pr.new/@tanstack/ai-grok@794

@tanstack/ai-groq

npm i https://pkg.pr.new/@tanstack/ai-groq@794

@tanstack/ai-isolate-cloudflare

npm i https://pkg.pr.new/@tanstack/ai-isolate-cloudflare@794

@tanstack/ai-isolate-node

npm i https://pkg.pr.new/@tanstack/ai-isolate-node@794

@tanstack/ai-isolate-quickjs

npm i https://pkg.pr.new/@tanstack/ai-isolate-quickjs@794

@tanstack/ai-mcp

npm i https://pkg.pr.new/@tanstack/ai-mcp@794

@tanstack/ai-ollama

npm i https://pkg.pr.new/@tanstack/ai-ollama@794

@tanstack/ai-openai

npm i https://pkg.pr.new/@tanstack/ai-openai@794

@tanstack/ai-openrouter

npm i https://pkg.pr.new/@tanstack/ai-openrouter@794

@tanstack/ai-preact

npm i https://pkg.pr.new/@tanstack/ai-preact@794

@tanstack/ai-react

npm i https://pkg.pr.new/@tanstack/ai-react@794

@tanstack/ai-react-ui

npm i https://pkg.pr.new/@tanstack/ai-react-ui@794

@tanstack/ai-solid

npm i https://pkg.pr.new/@tanstack/ai-solid@794

@tanstack/ai-solid-ui

npm i https://pkg.pr.new/@tanstack/ai-solid-ui@794

@tanstack/ai-svelte

npm i https://pkg.pr.new/@tanstack/ai-svelte@794

@tanstack/ai-utils

npm i https://pkg.pr.new/@tanstack/ai-utils@794

@tanstack/ai-vue

npm i https://pkg.pr.new/@tanstack/ai-vue@794

@tanstack/ai-vue-ui

npm i https://pkg.pr.new/@tanstack/ai-vue-ui@794

@tanstack/openai-base

npm i https://pkg.pr.new/@tanstack/openai-base@794

@tanstack/preact-ai-devtools

npm i https://pkg.pr.new/@tanstack/preact-ai-devtools@794

@tanstack/react-ai-devtools

npm i https://pkg.pr.new/@tanstack/react-ai-devtools@794

@tanstack/solid-ai-devtools

npm i https://pkg.pr.new/@tanstack/solid-ai-devtools@794

commit: 8873b33