fix(agent-core-v2): align externalHooks context injection and UserPromptSubmit payloads by luyao10282 · Pull Request #4256 · MoonshotAI/kimi-code

luyao10282 · 2026-10-16T08:08:47Z

Requirement or Bug

Resolve #4255

Bug Reproduction Steps

Configure a UserPromptSubmit or PreToolUse hook in ~/.kimi-code/config.toml that outputs {"hookSpecificOutput":{"hookEventName":"UserPromptSubmit","additionalContext":"context text"}}.
Send a prompt or invoke a slash skill (/my-skill arg1):
- On normal prompts, additionalContext is silently dropped because HookSpecificOutputSchema omits additionalContext and structuredOutput === true suppresses the stdout fallback.
- On slash skills (origin.kind === 'skill_activation'), UserPromptSubmit is skipped completely, and on normal prompts parsed.prompt on stdin is a ContentPart[] array rather than a string.
- On PreToolUse and PostToolUse, additionalContext and updatedInput are ignored.

Root Cause

HookSpecificOutputSchema in packages/agent-core-v2/src/features/externalHooks/internal/runHook.ts only extracted message, permissionDecision, and permissionDecisionReason, omitting additionalContext, systemMessage, and updatedInput.
runPromptSubmitHook in agentExternalHooksService.ts gated on origin.kind === 'user' and passed raw ContentPart[] as prompt.
runPreToolUse and notifyPostToolUse did not collect structured hook messages or updatedInput, and ExternalHooksRunnerService did not subscribe to config.onDidSectionChange.

This is a fundamental fix in packages/agent-core-v2.

Code Changes

Extended HookSpecificOutputSchema and HookJsonOutputSchema in internal/runHook.ts to parse additionalContext, systemMessage, and updatedInput.
Updated runPromptSubmitHook to run on user-slash skill_activation and plugin_command origins and serialize prompt as a string via turnPromptText.
Injected structured additionalContext / message from SessionStart, PreToolUse, and PostToolUse as <system-reminder><hook_result ...> messages (while ignoring noisy plain stdout on tool hooks), and wired updatedInput into BeforeToolExecuteEvent / AgentToolExecutorService with schema re-validation and resolveExecution re-evaluation.
Subscribed ExternalHooksRunnerService to config.onDidSectionChange for HOOKS_SECTION.

Behavior Changes and Affected Users

Behavior	Before	After	Who relies on the old behavior	Escape hatch
`hookSpecificOutput.additionalContext` in hook JSON output	Silently dropped (`structuredOutput: true` with `message: undefined`)	Extracted as hook message on `UserPromptSubmit`, `SessionStart`, `PreToolUse`, and `PostToolUse`	None (previously a no-op bug)	Omit `additionalContext` from hook JSON output
`UserPromptSubmit` on `user-slash` skill/plugin commands & `prompt` stdin field	Skipped on slash skills; `prompt` serialized as `ContentPart[]` array	Runs on `user-slash` skill/plugin activations; `prompt` serialized as string	None	Use `matcher` on `UserPromptSubmit` to filter slash commands
`PreToolUse` `updatedInput` and `PostToolUse` execution	`updatedInput` ignored; `PostToolUse` fire-and-forget without `tool_response`	`updatedInput` updates and re-validates tool args; `PostToolUse` awaited with `tool_response`	Hooks writing incidental plain `stdout` on tool events remain unaffected (`allowPlainStdout: false`)	N/A

Affected modules: packages/agent-core-v2/src/features/externalHooks, packages/agent-core-v2/src/agent/toolExecutor.
Test coverage: packages/agent-core-v2/test/features/externalHooks/integration.test.ts (142/142 tests passing).

Checklist

I have read the CONTRIBUTING document.
I have linked a related issue (external PRs: issue must have a maintainer's /approve).
I have added tests that prove my feature works.
The behavior-change table above is complete, and every removed behavior or flipped default is named in the changeset and either has an escape hatch or was explicitly approved by a maintainer in this PR.
Ran gen-changesets skill, or this PR needs no changeset.
Ran gen-docs skill, or this PR needs no doc update.

…mptSubmit payloads - Parse hookSpecificOutput.additionalContext, systemMessage, and updatedInput in runHook - Fire UserPromptSubmit on user-slash skill and plugin-command activations and pass prompt as string instead of ContentPart[] - Inject additionalContext on SessionStart, PreToolUse, and PostToolUse as system-reminder hook_result messages while ignoring noisy plain stdout on tool hooks - Support updatedInput mutation and re-validation in BeforeToolExecuteEvent / AgentToolExecutorService - Await PostToolUse and include tool_response in stdin payload - Reload hooks index when the hooks config section changes

changeset-bot · 2026-10-16T08:08:53Z

🦋 Changeset detected

Latest commit: fca6815

The changes in this PR will be included in the next version bump.

This PR includes changesets to release 2 packages

Name	Type
@moonshot-ai/kimi-code	Patch
@moonshot-ai/kimi-code-sdk	Patch

Not sure what this means? Click here to learn what changesets are.

Click here if you're a maintainer who wants to add another changeset to this PR

chatgpt-codex-connector · 2026-10-16T08:08:57Z

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-10-16T12:45:01.257995Z	`fca6815`	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

Copilot

🟡 Changes recommended

Rewritten executions retain stale authorization, and argument propagation, prompt handling, and context preservation defects remain.

5 open findings

What changed in this PR

Addresses #4255 by improving external-hook context injection and prompt payloads in the agent engine.

Changes:

Parses structured hook context and tool-input updates.
Extends prompt hooks to user-invoked slash commands.
Adds hook configuration reloads and integration coverage.

File	Description
`packages/agent-core-v2/test/features/externalHooks/integration.test.ts`	Tests slash prompts and tool-hook outputs.
`packages/agent-core-v2/src/features/externalHooks/session/sessionExternalHooksService.ts`	Queues SessionStart context.
`packages/agent-core-v2/src/features/externalHooks/session/sessionExternalHooks.ts`	Exposes queued context retrieval.
`packages/agent-core-v2/src/features/externalHooks/internal/userPrompt.ts`	Extracts messages and input updates.
`packages/agent-core-v2/src/features/externalHooks/internal/types.ts`	Adds updated-input result data.
`packages/agent-core-v2/src/features/externalHooks/internal/runHook.ts`	Parses additional structured output fields.
`packages/agent-core-v2/src/features/externalHooks/app/externalHooksRunnerService.ts`	Reloads hooks after configuration changes.
`packages/agent-core-v2/src/features/externalHooks/agent/agentExternalHooksService.ts`	Integrates context, input updates, and slash prompts.
`packages/agent-core-v2/src/agent/toolExecutor/toolHooks.ts`	Extends argument-update contracts.
`packages/agent-core-v2/src/agent/toolExecutor/toolExecutorService.ts`	Revalidates and resolves rewritten executions.
`packages/agent-core-v2/src/agent/toolExecutor/beforeToolExecuteEvent.ts`	Carries argument-update decisions.
`.changeset/hooks-context-and-prompt-payload.md`	Records user-facing hook changes.

🧠 Review effort: Balanced

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

+      try {
+        execution = await call.tool.resolveExecution(call.args);


    signal.throwIfAborted();
    const results = await this.runner.trigger('UserPromptSubmit', {
-      matcherValue: input,
+      matcherValue: promptText !== undefined ? promptText : input,


+    if (decision?.updatedArgs !== undefined) {
+      const validationError = validateExecutableToolArgs(call.tool, call.args);


      sessionId: this.sessionContext.sessionId,
-      inputData: this.withSessionFacts({ prompt: input, isSteer: ctx.isSteer }),
+      inputData: this.withSessionFacts({
+        prompt: promptText ?? input,


+        message ??
+        systemMessage ??
+        additionalContext ??
+        hookSpecificOutput?.message ??
+        hookSpecificOutput?.additionalContext,


chatgpt-codex-connector

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4498f658ae

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

chatgpt-codex-connector · 2026-10-16T08:17:45Z

+      try {
+        execution = await call.tool.resolveExecution(call.args);


Re-run permission checks for updated tool executions

When a PreToolUse hook changes a path-bearing input (for example, changing Write.path from a workspace file to an absolute outside-workspace or .git path), this new resolveExecution happens only after fireBeforeExecute has already run every permission listener against the original execution.accesses. The updated execution is then scheduled without re-running those policies, so it can execute under an approval or UI description computed for a different target; re-evaluate permissions against the updated execution before accepting the earlier decision.

AGENTS.md reference: AGENTS.md:L122-L124

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-10-16T08:17:45Z

+      ? `/${origin.commandName} ${args}`
+      : `/${origin.commandName}`;


Include the plugin namespace in submitted command text

For a user slash plugin command such as /demo:deploy prod, AgentPluginCommandService stores pluginId: "demo" and commandName: "deploy", and promptMetadataTextFromPluginCommand renders the canonical prompt as /demo:deploy prod; this branch instead sends /deploy prod as the matcher and stdin prompt. Hooks matching the namespaced command therefore do not run, while hooks that do run see a command the user never entered, so construct this text from both origin.pluginId and origin.commandName.

AGENTS.md reference: AGENTS.md:L114-L116

Useful? React with 👍 / 👎.

chatgpt-codex-connector · 2026-10-16T08:17:45Z

+      inputData: this.withSessionFacts({
+        prompt: promptText ?? input,
+        isSteer: ctx.isSteer,


Keep prompt payloads string-valued for media-only submissions

When kap-server accepts an image/video-only or empty-text prompt, turnPromptText returns undefined, so this fallback puts the original ContentPart[] back into stdin even though textual prompts now use a string. A UserPromptSubmit hook written to the new string payload contract will fail or silently allow those valid submissions; use a string fallback such as the empty string instead.

AGENTS.md reference: AGENTS.md:L114-L116

Useful? React with 👍 / 👎.

…matcher and hook context

Copilot AI balanced review requested due to automatic review settings October 16, 2026 08:08

Copilot started reviewing on behalf of luyao10282 October 16, 2026 08:09

View session

Copilot AI reviewed Oct 16, 2026

View reviewed changes

chatgpt-codex-connector Bot reviewed Oct 16, 2026

View reviewed changes

fix(hooks): re-authorize updated tool executions and preserve prompt …

fca6815

…matcher and hook context

Navigation Menu

Repository navigation

fix(agent-core-v2): align externalHooks context injection and UserPromptSubmit payloads - #4256

fix(agent-core-v2): align externalHooks context injection and UserPromptSubmit payloads#4256
luyao10282 wants to merge 2 commits into
MoonshotAI:mainfrom
luyao10282:feat/cc-hooks-compat

luyao10282 commented Oct 16, 2026

Uh oh!

changeset-bot Bot commented Oct 16, 2026 •

edited

Loading

Uh oh!

chatgpt-codex-connector Bot commented Oct 16, 2026 •

edited

Loading

Uh oh!

Copilot AI left a comment

Uh oh!

chatgpt-codex-connector Bot left a comment

Uh oh!

chatgpt-codex-connector Bot Oct 16, 2026

Uh oh!

chatgpt-codex-connector Bot Oct 16, 2026

Uh oh!

chatgpt-codex-connector Bot Oct 16, 2026

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		try {
		execution = await call.tool.resolveExecution(call.args);

		if (decision?.updatedArgs !== undefined) {
		const validationError = validateExecutableToolArgs(call.tool, call.args);

		? `/${origin.commandName} ${args}`
		: `/${origin.commandName}`;

Conversation

luyao10282 commented Oct 16, 2026

Requirement or Bug

Bug Reproduction Steps

Root Cause

Code Changes

Behavior Changes and Affected Users

Checklist

Uh oh!

changeset-bot Bot commented Oct 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🦋 Changeset detected

Uh oh!

chatgpt-codex-connector Bot commented Oct 16, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codex Review Summary

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

chatgpt-codex-connector Bot left a comment

Choose a reason for hiding this comment

💡 Codex Review

Uh oh!

chatgpt-codex-connector Bot Oct 16, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Oct 16, 2026

Choose a reason for hiding this comment

Uh oh!

chatgpt-codex-connector Bot Oct 16, 2026

Choose a reason for hiding this comment

Uh oh!

This branch has not been deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

changeset-bot Bot commented Oct 16, 2026 •

edited

Loading

chatgpt-codex-connector Bot commented Oct 16, 2026 •

edited

Loading