feat(hooks,skills): add gateguard fact-forcing pre-action gate

A PreToolUse hook that forces Claude to investigate before editing.
Instead of self-evaluation ("are you sure?"), it demands concrete facts:
importers, public API, data schemas, user instruction.

A/B tested: +2.25 quality points (9.0 vs 6.75) across two independent tasks.

- scripts/hooks/gateguard-fact-force.js — standalone Node.js hook
- skills/gateguard/SKILL.md — skill documentation
- hooks/hooks.json — PreToolUse entries for Edit|Write and Bash

Full package with config: pip install gateguard-ai
Repo: https://github.com/zunoworks/gateguard

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
seto
2026-04-12 11:41:33 +09:00
parent 125d5e6199
commit 5a03922934
3 changed files with 357 additions and 0 deletions

View File

@@ -126,6 +126,30 @@
], ],
"description": "Check MCP server health before MCP tool execution and block unhealthy MCP calls", "description": "Check MCP server health before MCP tool execution and block unhealthy MCP calls",
"id": "pre:mcp-health-check" "id": "pre:mcp-health-check"
},
{
"matcher": "Edit|Write",
"hooks": [
{
"type": "command",
"command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/hooks/gateguard-fact-force.js\"",
"timeout": 5
}
],
"description": "Fact-forcing gate: block first Edit/Write per file and demand investigation (importers, data schemas, user instruction) before allowing",
"id": "pre:edit-write:gateguard-fact-force"
},
{
"matcher": "Bash",
"hooks": [
{
"type": "command",
"command": "node \"${CLAUDE_PLUGIN_ROOT}/scripts/hooks/gateguard-fact-force.js\"",
"timeout": 5
}
],
"description": "Fact-forcing gate: block destructive Bash commands and demand rollback plan; quote user instruction on first Bash per session",
"id": "pre:bash:gateguard-fact-force"
} }
], ],
"PreCompact": [ "PreCompact": [

View File

@@ -0,0 +1,216 @@
#!/usr/bin/env node
/**
* PreToolUse Hook: GateGuard Fact-Forcing Gate
*
* Forces Claude to investigate before editing files or running commands.
* Instead of asking "are you sure?" (which LLMs always answer "yes"),
* this hook demands concrete facts: importers, public API, data schemas.
*
* The act of investigation creates awareness that self-evaluation never did.
*
* Gates:
* - Edit/Write: list importers, affected API, verify data schemas, quote instruction
* - Bash (destructive): list targets, rollback plan, quote instruction
* - Bash (routine): quote current instruction (once per session)
*
* Exit codes:
* 0 - Allow (gate already passed for this target)
* 2 - Block (force investigation first)
*
* Cross-platform (Windows, macOS, Linux).
*
* Full package with config support: pip install gateguard-ai
* Repo: https://github.com/zunoworks/gateguard
*/
'use strict';
const fs = require('fs');
const path = require('path');
const MAX_STDIN = 1024 * 1024;
// Session state file for tracking which files have been gated
const STATE_DIR = path.join(process.env.HOME || process.env.USERPROFILE || '/tmp', '.gateguard');
const STATE_FILE = path.join(STATE_DIR, '.session_state.json');
const DESTRUCTIVE_BASH = /\b(rm\s+-rf|git\s+reset\s+--hard|git\s+checkout\s+--|git\s+clean\s+-f|drop\s+table|delete\s+from|truncate|git\s+push\s+--force|dd\s+if=)\b/i;
// --- State management ---
function loadState() {
try {
if (fs.existsSync(STATE_FILE)) {
return JSON.parse(fs.readFileSync(STATE_FILE, 'utf8'));
}
} catch (_) { /* ignore */ }
return { checked: [], read_files: [] };
}
function saveState(state) {
try {
fs.mkdirSync(STATE_DIR, { recursive: true });
fs.writeFileSync(STATE_FILE, JSON.stringify(state, null, 2), 'utf8');
} catch (_) { /* ignore */ }
}
function markChecked(key) {
const state = loadState();
if (!state.checked.includes(key)) {
state.checked.push(key);
saveState(state);
}
}
function isChecked(key) {
const state = loadState();
return state.checked.includes(key);
}
// --- Sanitize file path against injection ---
function sanitizePath(filePath) {
return filePath.replace(/[\n\r]/g, ' ').trim().slice(0, 500);
}
// --- Gate messages ---
function editGateMsg(filePath) {
const safe = sanitizePath(filePath);
return [
'[Fact-Forcing Gate]',
'',
`Before editing ${safe}, present these facts:`,
'',
'1. List ALL files that import/require this file (use Grep)',
'2. List the public functions/classes affected by this change',
'3. If this file reads/writes data files, cat one real record and show actual field names, structure, and date format',
'4. Quote the user\'s current instruction verbatim',
'',
'Present the facts, then retry the same operation.'
].join('\n');
}
function writeGateMsg(filePath) {
const safe = sanitizePath(filePath);
return [
'[Fact-Forcing Gate]',
'',
`Before creating ${safe}, present these facts:`,
'',
'1. Name the file(s) and line(s) that will call this new file',
'2. Confirm no existing file serves the same purpose (use Glob)',
'3. If this file reads/writes data files, cat one real record and show actual field names, structure, and date format',
'4. Quote the user\'s current instruction verbatim',
'',
'Present the facts, then retry the same operation.'
].join('\n');
}
function destructiveBashMsg() {
return [
'[Fact-Forcing Gate]',
'',
'Destructive command detected. Before running, present:',
'',
'1. List all files/data this command will modify or delete',
'2. Write a one-line rollback procedure',
'3. Quote the user\'s current instruction verbatim',
'',
'Present the facts, then retry the same operation.'
].join('\n');
}
function routineBashMsg() {
return [
'[Fact-Forcing Gate]',
'',
'Quote the user\'s current instruction verbatim.',
'Then retry the same operation.'
].join('\n');
}
// --- Output helpers ---
function deny(reason) {
const output = {
hookSpecificOutput: {
hookEventName: 'PreToolUse',
permissionDecision: 'deny',
permissionDecisionReason: reason
}
};
process.stdout.write(JSON.stringify(output));
process.exit(0);
}
function allow() {
// Output nothing = allow
process.exit(0);
}
// --- Main ---
function main() {
let raw = '';
try {
raw = fs.readFileSync(0, 'utf8').slice(0, MAX_STDIN);
} catch (_) {
allow();
return;
}
let data;
try {
data = JSON.parse(raw);
} catch (_) {
allow();
return;
}
const toolName = data.tool_name || '';
const toolInput = data.tool_input || {};
if (toolName === 'Edit' || toolName === 'Write') {
const filePath = toolInput.file_path || '';
if (!filePath) {
allow();
return;
}
// Gate: first action per file
if (!isChecked(filePath)) {
markChecked(filePath);
const msg = toolName === 'Edit' ? editGateMsg(filePath) : writeGateMsg(filePath);
deny(msg);
return;
}
allow();
return;
}
if (toolName === 'Bash') {
const command = toolInput.command || '';
// Destructive commands: always gate
if (DESTRUCTIVE_BASH.test(command)) {
deny(destructiveBashMsg());
return;
}
// Routine bash: once per session
if (!isChecked('__bash_session__')) {
markChecked('__bash_session__');
deny(routineBashMsg());
return;
}
allow();
return;
}
allow();
}
main();

117
skills/gateguard/SKILL.md Normal file
View File

@@ -0,0 +1,117 @@
---
name: gateguard
description: Fact-forcing gate that blocks Edit/Write/Bash and demands concrete investigation (importers, data schemas, user instruction) before allowing the action. Measurably improves output quality by +2.25 points vs ungated agents.
origin: community
---
# GateGuard — Fact-Forcing Pre-Action Gate
A PreToolUse hook that forces Claude to investigate before editing. Instead of self-evaluation ("are you sure?"), it demands concrete facts. The act of investigation creates awareness that self-evaluation never did.
## When to Activate
- Working on any codebase where file edits affect multiple modules
- Projects with data files that have specific schemas or date formats
- Teams where AI-generated code must match existing patterns
- Any workflow where Claude tends to guess instead of investigating
## Core Concept
LLM self-evaluation doesn't work. Ask "did you violate any policies?" and the answer is always "no." This is verified experimentally.
But asking "list every file that imports this module" forces the LLM to run Grep and Read. The investigation itself creates context that changes the output.
**Three-stage gate:**
```
1. DENY — block the first Edit/Write/Bash attempt
2. FORCE — tell the model exactly which facts to gather
3. ALLOW — permit retry after facts are presented
```
No competitor does all three. Most stop at deny.
## Evidence
Two independent A/B tests, identical agents, same task:
| Task | Gated | Ungated | Gap |
| --- | --- | --- | --- |
| Analytics module | 8.0/10 | 6.5/10 | +1.5 |
| Webhook validator | 10.0/10 | 7.0/10 | +3.0 |
| **Average** | **9.0** | **6.75** | **+2.25** |
Both agents produce code that runs and passes tests. The difference is design depth.
## Gate Types
### Edit Gate (first edit per file)
```
Before editing {file_path}, present these facts:
1. List ALL files that import/require this file (use Grep)
2. List the public functions/classes affected by this change
3. If this file reads/writes data files, cat one real record
and show actual field names, structure, and date format
4. Quote the user's current instruction verbatim
```
### Write Gate (first new file creation)
```
Before creating {file_path}, present these facts:
1. Name the file(s) and line(s) that will call this new file
2. Confirm no existing file serves the same purpose (use Glob)
3. If this file reads/writes data files, cat one real record
4. Quote the user's current instruction verbatim
```
### Destructive Bash Gate (every destructive command)
Triggers on: `rm -rf`, `git reset --hard`, `git push --force`, `drop table`, etc.
```
1. List all files/data this command will modify or delete
2. Write a one-line rollback procedure
3. Quote the user's current instruction verbatim
```
### Routine Bash Gate (once per session)
```
Quote the user's current instruction verbatim.
```
## Quick Start
### Option A: Use the ECC hook (zero install)
The hook at `scripts/hooks/gateguard-fact-force.js` is included in this plugin. Enable it via hooks.json.
### Option B: Full package with config
```bash
pip install gateguard-ai
gateguard init
```
This adds `.gateguard.yml` for per-project configuration (custom messages, ignore paths, gate toggles).
## Anti-Patterns
- **Don't use self-evaluation instead.** "Are you sure?" always gets "yes." This is experimentally verified.
- **Don't skip the data schema check.** Both A/B test agents assumed ISO-8601 dates when real data used `%Y/%m/%d %H:%M`. Checking one real record prevents this entire class of bugs.
- **Don't gate every single Bash command.** Routine bash gates once per session. Destructive bash gates every time. This balance avoids slowdown while catching real risks.
## Best Practices
- Let the gate fire naturally. Don't try to pre-answer the gate questions — the investigation itself is what improves quality.
- Customize gate messages for your domain. If your project has specific conventions, add them to the gate prompts.
- Use `.gateguard.yml` to ignore paths like `.venv/`, `node_modules/`, `.git/`.
## Related Skills
- `safety-guard` — Runtime safety checks (complementary, not overlapping)
- `code-reviewer` — Post-edit review (GateGuard is pre-edit investigation)