[BUG] pre_tool_use and post_tool_use hooks not firing

Open pm0code opened this issue 6 days ago • 1 comments

Preflight Checklist

[x] I have searched existing issues and this hasn't been reported yet
[x] This is a single bug report (please file separate reports for different bugs)
[x] I am using the latest version of Claude Code

What's Wrong?

Claude Code Bug Report: `pre_tool_use` and `post_tool_use` Hooks Not Firing

Date: 2025-12-26 Claude Code Version: 2.0.76 Platform: Windows 11 x64

Summary

Hookify hooks configured with event: pre_tool_use and event: post_tool_use are not triggering in Claude Code sessions, despite being properly configured with enabled: true and valid matchers.

Expected Behavior

When a hook is configured with:

---
name: mandatory-thinking-before-action
enabled: true
event: pre_tool_use
action: block
matcher: Grep|Glob|Read|Write|Edit|Bash|WebSearch|WebFetch|Task
---

Expected: Claude should be blocked from using any matching tools (Grep, Glob, Read, Write, Edit, Bash, WebSearch, WebFetch, Task) until the hook conditions are satisfied.

Actual Behavior

Actual: Claude freely uses all matching tools without any blocking, warnings, or acknowledgment that the hook exists.

Evidence

Hook Configuration (Verified)

File: D:\[...]\.claude\hookify.mandatory-thinking-before-action.local.md

---
name: mandatory-thinking-before-action
enabled: true
event: pre_tool_use
action: block
matcher: Grep|Glob|Read|Write|Edit|Bash|WebSearch|WebFetch|Task
---

## BLOCKED - MANDATORY THINKING BEFORE ANY ACTION

**Protocol v10.0.0 - ZERO TOLERANCE FOR REACTIVE BEHAVIOR**

### YOU CANNOT USE THIS TOOL UNTIL YOU HAVE:

**1. ARTICULATED THE PROBLEM** (in your response text)
**2. SHOWN YOUR THINKING** (in your response text)
**3. PROPOSED A PLAN** (in your response text)
**4. ASKED FOR APPROVAL** (for non-trivial tasks)

Tool Usage That Should Have Been Blocked

In the same session where the hook is enabled, Claude successfully executed:

Glob - .claude/hookify*.md (line 82216)
Read - Multiple reads of hookify config files (lines 82718-107795)
Bash - Multiple bash commands (lines 100766, 101838, etc.)
TodoWrite - Task list updates (line 82216)

None of these tool calls were blocked, warned, or acknowledged by the hook system.

Hook System IS Working for Other Event Types

Working hooks:

event: file hooks ✅ (detect file pattern changes)
event: bash hooks ✅ (detect command patterns)
event: stop hooks ✅ (completion validation)
event: UserPromptSubmit ✅ (shows "hook success" messages)

NOT Working hooks:

event: pre_tool_use ❌ (no blocking, no warnings)
event: post_tool_use ❌ (no verification prompts)

System Messages Confirming Other Hooks Work

<system-reminder>
UserPromptSubmit hook success: Success
</system-reminder>
<system-reminder>
UserPromptSubmit:Callback hook success: Success
</system-reminder>

These messages appear consistently, proving the hookify system IS active, but only for certain event types.

Reproduction Steps

Create a hookify hook file in .claude/ directory:

---
name: test-pre-tool-use-hook
enabled: true
event: pre_tool_use
action: block
matcher: Read|Glob
---

## TEST HOOK - YOU SHOULD BE BLOCKED

This hook should prevent Read and Glob tool usage.

Start a Claude Code session in that directory
Ask Claude to read a file or search for files
Expected: Claude is blocked and cannot use Read/Glob until hook conditions satisfied
Actual: Claude freely uses Read and Glob without acknowledgment of the hook

Impact

Critical for APEX-v2 Project

This bug breaks Protocol v10.0.0 enforcement in the APEX-v2 trading system project, which relies on pre_tool_use hooks to:

Prevent reactive coding - Force ANALYSIS→THINKING→PLAN→APPROVAL workflow before tool usage
Enforce holistic thinking - Block tools until full system impact is considered
Prevent costly mistakes - Require approval before code changes in financial systems

Examples of Broken Workflows

Hook: mandatory-thinking-before-action.local.md Purpose: Block ALL tool usage until Claude articulates problem, shows thinking, and gets approval Status: ❌ Not working - Claude uses tools immediately without planning

Hook: post-action-verification.local.md Purpose: After Write/Edit/Bash, verify changes had intended effect Status: ❌ Not working - No verification prompts appear

Workaround

Currently using manual compliance: Developer reminds Claude to follow the workflow before tool usage. This is error-prone and defeats the purpose of automated enforcement.

Diagnostic Information

Hook Files Present

$ ls .claude/hookify*.md
hookify.[...]-aiml-models-registry.local.md
hookify.[...]-completion-validation.local.md
hookify.[...]-database-operations.local.md
hookify.[...]-financial-precision.local.md
hookify.[...]-gpu-operations.local.md
hookify.[...]-protocol-violations.local.md
hookify.[...]-service-hardening.local.md
hookify.[...]-winui3-builds.local.md
hookify.financial-precision-check.local.md
hookify.mandatory-thinking-before-action.local.md  # <-- THIS ONE
hookify.post-action-verification.local.md          # <-- THIS ONE
hookify.require-agent-planning.local.md
hookify.require-analysis-before-code.local.md
hookify.ui-chain-verification.local.md

Enabled Hooks Count

$ grep -l "enabled: true" .claude/hookify*.md | wc -l
10

Pre-Tool-Use Hooks (Should Be Working)

$ grep -l "event: pre_tool_use" .claude/hookify*.md
hookify.mandatory-thinking-before-action.local.md  # enabled: true
hookify.require-agent-planning.local.md            # enabled: false (deprecated)
hookify.require-analysis-before-code.local.md      # enabled: false (deprecated)

Only 1 active pre_tool_use hook, and it's not firing.

Questions for Claude Code Team

Is pre_tool_use event implemented in the current Claude Code version?
Is there a different event name that should be used for pre-tool blocking?
Are there any known limitations or requirements for pre_tool_use hooks?
What is the roadmap for full hookify event support?

Suggested Fix

If pre_tool_use is not yet implemented:

Document which events ARE supported in hookify documentation
Add pre_tool_use and post_tool_use support to match documentation examples
Ensure action: block actually prevents tool execution (not just warnings)

Additional Context

This is blocking a critical quality enforcement mechanism for a financial trading system development project. The project has 14 hookify hooks configured to enforce:

Decimal precision for money (no float/double)
Database operation safety (architect consultation required)
GPU operation governance
AI/ML model registry compliance
Protocol v10.0.0 zero-tolerance policies

Without working pre_tool_use hooks, these enforcement mechanisms are non-functional.

What Should Happen?

Claude Code Bug Report: `pre_tool_use` and `post_tool_use` Hooks Not Firing

Date: 2025-12-26 Claude Code Version: 2.0.76 Platform: Windows 11 x64

Summary

Hookify hooks configured with event: pre_tool_use and event: post_tool_use are not triggering in Claude Code sessions, despite being properly configured with enabled: true and valid matchers.

Expected Behavior

When a hook is configured with:

---
name: mandatory-thinking-before-action
enabled: true
event: pre_tool_use
action: block
matcher: Grep|Glob|Read|Write|Edit|Bash|WebSearch|WebFetch|Task
---

Expected: Claude should be blocked from using any matching tools (Grep, Glob, Read, Write, Edit, Bash, WebSearch, WebFetch, Task) until the hook conditions are satisfied.

Actual Behavior

Actual: Claude freely uses all matching tools without any blocking, warnings, or acknowledgment that the hook exists.

Evidence

Hook Configuration (Verified)

File: D:\[...]\.claude\hookify.mandatory-thinking-before-action.local.md

---
name: mandatory-thinking-before-action
enabled: true
event: pre_tool_use
action: block
matcher: Grep|Glob|Read|Write|Edit|Bash|WebSearch|WebFetch|Task
---

## BLOCKED - MANDATORY THINKING BEFORE ANY ACTION

**Protocol v10.0.0 - ZERO TOLERANCE FOR REACTIVE BEHAVIOR**

### YOU CANNOT USE THIS TOOL UNTIL YOU HAVE:

**1. ARTICULATED THE PROBLEM** (in your response text)
**2. SHOWN YOUR THINKING** (in your response text)
**3. PROPOSED A PLAN** (in your response text)
**4. ASKED FOR APPROVAL** (for non-trivial tasks)

Tool Usage That Should Have Been Blocked

In the same session where the hook is enabled, Claude successfully executed:

Glob - .claude/hookify*.md (line 82216)
Read - Multiple reads of hookify config files (lines 82718-107795)
Bash - Multiple bash commands (lines 100766, 101838, etc.)
TodoWrite - Task list updates (line 82216)

None of these tool calls were blocked, warned, or acknowledged by the hook system.

Hook System IS Working for Other Event Types

Working hooks:

event: file hooks ✅ (detect file pattern changes)
event: bash hooks ✅ (detect command patterns)
event: stop hooks ✅ (completion validation)
event: UserPromptSubmit ✅ (shows "hook success" messages)

NOT Working hooks:

event: pre_tool_use ❌ (no blocking, no warnings)
event: post_tool_use ❌ (no verification prompts)

System Messages Confirming Other Hooks Work

<system-reminder>
UserPromptSubmit hook success: Success
</system-reminder>
<system-reminder>
UserPromptSubmit:Callback hook success: Success
</system-reminder>

These messages appear consistently, proving the hookify system IS active, but only for certain event types.

Reproduction Steps

Create a hookify hook file in .claude/ directory:

---
name: test-pre-tool-use-hook
enabled: true
event: pre_tool_use
action: block
matcher: Read|Glob
---

## TEST HOOK - YOU SHOULD BE BLOCKED

This hook should prevent Read and Glob tool usage.

Start a Claude Code session in that directory
Ask Claude to read a file or search for files
Expected: Claude is blocked and cannot use Read/Glob until hook conditions satisfied
Actual: Claude freely uses Read and Glob without acknowledgment of the hook

Impact

Critical for APEX-v2 Project

This bug breaks Protocol v10.0.0 enforcement in the APEX-v2 trading system project, which relies on pre_tool_use hooks to:

Prevent reactive coding - Force ANALYSIS→THINKING→PLAN→APPROVAL workflow before tool usage
Enforce holistic thinking - Block tools until full system impact is considered
Prevent costly mistakes - Require approval before code changes in financial systems

Examples of Broken Workflows

Hook: post-action-verification.local.md Purpose: After Write/Edit/Bash, verify changes had intended effect Status: ❌ Not working - No verification prompts appear

Workaround

Currently using manual compliance: Developer reminds Claude to follow the workflow before tool usage. This is error-prone and defeats the purpose of automated enforcement.

Diagnostic Information

Hook Files Present

$ ls .claude/hookify*.md
hookify.[...]-aiml-models-registry.local.md
hookify.[...]-completion-validation.local.md
hookify.[...]-database-operations.local.md
hookify.[...]-financial-precision.local.md
hookify.[...]-gpu-operations.local.md
hookify.[...]-protocol-violations.local.md
hookify.[...]-service-hardening.local.md
hookify.[...]-winui3-builds.local.md
hookify.financial-precision-check.local.md
hookify.mandatory-thinking-before-action.local.md  # <-- THIS ONE
hookify.post-action-verification.local.md          # <-- THIS ONE
hookify.require-agent-planning.local.md
hookify.require-analysis-before-code.local.md
hookify.ui-chain-verification.local.md

Enabled Hooks Count

$ grep -l "enabled: true" .claude/hookify*.md | wc -l
10

Pre-Tool-Use Hooks (Should Be Working)

$ grep -l "event: pre_tool_use" .claude/hookify*.md
hookify.mandatory-thinking-before-action.local.md  # enabled: true
hookify.require-agent-planning.local.md            # enabled: false (deprecated)
hookify.require-analysis-before-code.local.md      # enabled: false (deprecated)

Only 1 active pre_tool_use hook, and it's not firing.

Questions for Claude Code Team

Is pre_tool_use event implemented in the current Claude Code version?
Is there a different event name that should be used for pre-tool blocking?
Are there any known limitations or requirements for pre_tool_use hooks?
What is the roadmap for full hookify event support?

Suggested Fix

If pre_tool_use is not yet implemented:

Document which events ARE supported in hookify documentation
Add pre_tool_use and post_tool_use support to match documentation examples
Ensure action: block actually prevents tool execution (not just warnings)

Additional Context

This is blocking a critical quality enforcement mechanism for a financial trading system development project. The project has 14 hookify hooks configured to enforce:

Decimal precision for money (no float/double)
Database operation safety (architect consultation required)
GPU operation governance
AI/ML model registry compliance
Protocol v10.0.0 zero-tolerance policies

Without working pre_tool_use hooks, these enforcement mechanisms are non-functional.

Error Messages/Logs

Steps to Reproduce

Claude Code Bug Report: `pre_tool_use` and `post_tool_use` Hooks Not Firing

Date: 2025-12-26 Claude Code Version: 2.0.76 Platform: Windows 11 x64

Summary

Hookify hooks configured with event: pre_tool_use and event: post_tool_use are not triggering in Claude Code sessions, despite being properly configured with enabled: true and valid matchers.

Expected Behavior

When a hook is configured with:

---
name: mandatory-thinking-before-action
enabled: true
event: pre_tool_use
action: block
matcher: Grep|Glob|Read|Write|Edit|Bash|WebSearch|WebFetch|Task
---

Expected: Claude should be blocked from using any matching tools (Grep, Glob, Read, Write, Edit, Bash, WebSearch, WebFetch, Task) until the hook conditions are satisfied.

Actual Behavior

Actual: Claude freely uses all matching tools without any blocking, warnings, or acknowledgment that the hook exists.

Evidence

Hook Configuration (Verified)

File: D:\[...]\.claude\hookify.mandatory-thinking-before-action.local.md

---
name: mandatory-thinking-before-action
enabled: true
event: pre_tool_use
action: block
matcher: Grep|Glob|Read|Write|Edit|Bash|WebSearch|WebFetch|Task
---

## BLOCKED - MANDATORY THINKING BEFORE ANY ACTION

**Protocol v10.0.0 - ZERO TOLERANCE FOR REACTIVE BEHAVIOR**

### YOU CANNOT USE THIS TOOL UNTIL YOU HAVE:

**1. ARTICULATED THE PROBLEM** (in your response text)
**2. SHOWN YOUR THINKING** (in your response text)
**3. PROPOSED A PLAN** (in your response text)
**4. ASKED FOR APPROVAL** (for non-trivial tasks)

Tool Usage That Should Have Been Blocked

In the same session where the hook is enabled, Claude successfully executed:

Glob - .claude/hookify*.md (line 82216)
Read - Multiple reads of hookify config files (lines 82718-107795)
Bash - Multiple bash commands (lines 100766, 101838, etc.)
TodoWrite - Task list updates (line 82216)

None of these tool calls were blocked, warned, or acknowledged by the hook system.

Hook System IS Working for Other Event Types

Working hooks:

event: file hooks ✅ (detect file pattern changes)
event: bash hooks ✅ (detect command patterns)
event: stop hooks ✅ (completion validation)
event: UserPromptSubmit ✅ (shows "hook success" messages)

NOT Working hooks:

event: pre_tool_use ❌ (no blocking, no warnings)
event: post_tool_use ❌ (no verification prompts)

System Messages Confirming Other Hooks Work

<system-reminder>
UserPromptSubmit hook success: Success
</system-reminder>
<system-reminder>
UserPromptSubmit:Callback hook success: Success
</system-reminder>

These messages appear consistently, proving the hookify system IS active, but only for certain event types.

Reproduction Steps

Create a hookify hook file in .claude/ directory:

---
name: test-pre-tool-use-hook
enabled: true
event: pre_tool_use
action: block
matcher: Read|Glob
---

## TEST HOOK - YOU SHOULD BE BLOCKED

This hook should prevent Read and Glob tool usage.

Start a Claude Code session in that directory
Ask Claude to read a file or search for files
Expected: Claude is blocked and cannot use Read/Glob until hook conditions satisfied
Actual: Claude freely uses Read and Glob without acknowledgment of the hook

Impact

Critical for APEX-v2 Project

This bug breaks Protocol v10.0.0 enforcement in the APEX-v2 trading system project, which relies on pre_tool_use hooks to:

Prevent reactive coding - Force ANALYSIS→THINKING→PLAN→APPROVAL workflow before tool usage
Enforce holistic thinking - Block tools until full system impact is considered
Prevent costly mistakes - Require approval before code changes in financial systems

Examples of Broken Workflows

Hook: post-action-verification.local.md Purpose: After Write/Edit/Bash, verify changes had intended effect Status: ❌ Not working - No verification prompts appear

Workaround

Currently using manual compliance: Developer reminds Claude to follow the workflow before tool usage. This is error-prone and defeats the purpose of automated enforcement.

Diagnostic Information

Hook Files Present

$ ls .claude/hookify*.md
hookify.[...]-aiml-models-registry.local.md
hookify.[...]-completion-validation.local.md
hookify.[...]-database-operations.local.md
hookify.[...]-financial-precision.local.md
hookify.[...]-gpu-operations.local.md
hookify.[...]-protocol-violations.local.md
hookify.[...]-service-hardening.local.md
hookify.[...]-winui3-builds.local.md
hookify.financial-precision-check.local.md
hookify.mandatory-thinking-before-action.local.md  # <-- THIS ONE
hookify.post-action-verification.local.md          # <-- THIS ONE
hookify.require-agent-planning.local.md
hookify.require-analysis-before-code.local.md
hookify.ui-chain-verification.local.md

Enabled Hooks Count

$ grep -l "enabled: true" .claude/hookify*.md | wc -l
10

Pre-Tool-Use Hooks (Should Be Working)

$ grep -l "event: pre_tool_use" .claude/hookify*.md
hookify.mandatory-thinking-before-action.local.md  # enabled: true
hookify.require-agent-planning.local.md            # enabled: false (deprecated)
hookify.require-analysis-before-code.local.md      # enabled: false (deprecated)

Only 1 active pre_tool_use hook, and it's not firing.

Questions for Claude Code Team

Is pre_tool_use event implemented in the current Claude Code version?
Is there a different event name that should be used for pre-tool blocking?
Are there any known limitations or requirements for pre_tool_use hooks?
What is the roadmap for full hookify event support?

Suggested Fix

If pre_tool_use is not yet implemented:

Document which events ARE supported in hookify documentation
Add pre_tool_use and post_tool_use support to match documentation examples
Ensure action: block actually prevents tool execution (not just warnings)

Additional Context

This is blocking a critical quality enforcement mechanism for a financial trading system development project. The project has 14 hookify hooks configured to enforce:

Decimal precision for money (no float/double)
Database operation safety (architect consultation required)
GPU operation governance
AI/ML model registry compliance
Protocol v10.0.0 zero-tolerance policies

Without working pre_tool_use hooks, these enforcement mechanisms are non-functional.

Claude Model

None

Is this a regression?

Yes, this worked in a previous version

Last Working Version

No response

Claude Code Version

2.0.76

Platform

Anthropic API

Operating System

Windows

Terminal/Shell

Terminal.app (macOS)

Additional Information

No response

Dec 26 '25 17:12 pm0code

[BUG] pre_tool_use and post_tool_use hooks not firing

Preflight Checklist

What's Wrong?

Claude Code Bug Report: pre_tool_use and post_tool_use Hooks Not Firing

Summary

Expected Behavior

Actual Behavior

Evidence

Hook Configuration (Verified)

Tool Usage That Should Have Been Blocked

Hook System IS Working for Other Event Types

System Messages Confirming Other Hooks Work

Reproduction Steps

Impact

Critical for APEX-v2 Project

Examples of Broken Workflows

Workaround

Diagnostic Information

Hook Files Present

Enabled Hooks Count

Pre-Tool-Use Hooks (Should Be Working)

Questions for Claude Code Team

Suggested Fix

Additional Context

What Should Happen?

Claude Code Bug Report: pre_tool_use and post_tool_use Hooks Not Firing

Summary

Expected Behavior

Actual Behavior

Evidence

Hook Configuration (Verified)

Tool Usage That Should Have Been Blocked

Hook System IS Working for Other Event Types

System Messages Confirming Other Hooks Work

Reproduction Steps

Impact

Critical for APEX-v2 Project

Examples of Broken Workflows

Workaround

Diagnostic Information

Hook Files Present

Enabled Hooks Count

Pre-Tool-Use Hooks (Should Be Working)

Questions for Claude Code Team

Suggested Fix

Additional Context

Error Messages/Logs

Steps to Reproduce

Claude Code Bug Report: pre_tool_use and post_tool_use Hooks Not Firing

Summary

Expected Behavior

Actual Behavior

Evidence

Hook Configuration (Verified)

Tool Usage That Should Have Been Blocked

Hook System IS Working for Other Event Types

System Messages Confirming Other Hooks Work

Reproduction Steps

Impact

Critical for APEX-v2 Project

Examples of Broken Workflows

Workaround

Diagnostic Information

Hook Files Present

Enabled Hooks Count

Pre-Tool-Use Hooks (Should Be Working)

Questions for Claude Code Team

Suggested Fix

Additional Context

Claude Model

Is this a regression?

Last Working Version

Claude Code Version

Platform

Operating System

Terminal/Shell

Additional Information

Claude Code Bug Report: `pre_tool_use` and `post_tool_use` Hooks Not Firing

Claude Code Bug Report: `pre_tool_use` and `post_tool_use` Hooks Not Firing

Claude Code Bug Report: `pre_tool_use` and `post_tool_use` Hooks Not Firing