← Search

GateGuard Fact-Forcing Pre-Action Gate
Claude Code 🧪 Install ran in our test sandbox · Nov 11, 2026 📌 Pinned version: commit 4eb71d9
Install test only, on the date shown. This does not guarantee the code is safe.
⚙ Needs: Claude Code with hooks support…

GateGuard Fact-Forcing Pre-Action Gate

A PreToolUse hook that blocks Claude's first Edit, Write, or Bash attempt until the agent presents concrete facts: every importer, the data schema, the verbatim instruction. A/B-tested at +2.25 quality points.

At a glance
What
A PreToolUse hook that blocks Claude's first Edit, Write, or Bash attempt until the agent presents concrete facts: every importer, the data schema, the verbatim instruction. A/B-tested at +2.25 quality points.
Cost
Free
Needs
Claude Code with hooks support; nothing else — no accounts or keys required. The gate itself is a hook script that ships with the project (scripts/hooks/gateguard-fact-force.js) and is switched on through hooks.json; this install gives you the skill file, and my report at the end covers turning the hook on.
Level
Beginner — Claude Code with hooks support…
Install
Copy the installer prompt below into your Muse — your agent does the rest.

Version:

Quick answer

GateGuard Fact-Forcing Pre-Action Gate is a free Muse skill and PreToolUse hook that blocks Claude's first Edit, Write or Bash attempt until the agent presents facts: importers, schema and verbatim instruction. Instead of asking the model to self-evaluate, which always gets yes, GateGuard runs a three-stage gate: DENY, FORCE fact gathering, then ALLOW, tested at plus 2.25 quality points.

⌁

Install

Copy the install package below, then paste it into Muse🧪 Install ran in our test sandbox · Nov 11, 2026📌 Pinned version: commit 4eb71d9

Before you start: tell me in a few lines what this skill does and what installing it involves, and wait for my go-ahead. Prerequisites: Claude Code with hooks support; nothing else — no accounts or keys required. The gate itself is a hook script that ships with the project (scripts/hooks/gateguard-fact-force.js) and is switched on through hooks.json; this install gives you the skill file, and my report at the end covers turning the hook on. Install "GateGuard Fact-Forcing Pre-Action Gate" for me. GateGuard is a guardrail for Claude Code: a PreToolUse hook that briefly blocks the first edit, new file or shell command on anything new, and demands a fixed set of concrete facts before it lets the work go ahead — never questions that adapt to the file or the change. The Edit/MultiEdit gate asks for four facts: every file that imports the target, the public functions or classes affected, the data file schema with redacted values, and the user's instruction quoted verbatim. The Write gate asks for four facts: the files and lines that will call the new file, proof no existing file serves the same purpose, the data file schema with redacted values, and the verbatim user instruction. The destructive Bash gate fires on every destructive command (rm -rf, git reset --hard, git push --force, drop table) and asks for three facts: everything the command will modify or delete, a one-line rollback procedure, and the verbatim user instruction. The routine Bash gate fires once per session and asks for two facts: the current request in one sentence, and what the command verifies or produces. Two independent A/B tests measured +2.25 average quality points, the difference showing in design depth, not just passing tests. In a parallel batch the first call is denied while siblings may already apply, so send dependent edits to an untouched file sequentially and re-read the file after a denial. I can exempt folders I trust in advance with the GATEGUARD_EXEMPT_GLOBS setting, cap how many times the gate can stop me per session with GATEGUARD_FACT_FORCE_MAX_DENIALS (destructive Bash stays gated), and I can turn the whole gate off for a session by starting with ECC_GATEGUARD=off. A full package is also available (pip install gateguard-ai, then gateguard init) with a .gateguard.yml per-project configuration file. Source: https://raw.githubusercontent.com/affaan-m/ECC/4eb71d92a39cab44ad40ac9d8a6a5ccb4029d6c2/skills/gateguard/SKILL.md (the hook script lives in the same project, in the scripts/hooks folder at its root) Pinned version: install the code at commit 4eb71d92a39cab44ad40ac9d8a6a5ccb4029d6c2 exactly (Skill Harbor tested this version). Do not substitute a newer version. If this commit is unreachable, stop and tell me. Updated November 17, 2026: Skill Harbor rewrote this install prompt because the previous version described a different, newer and much larger version of the skill than the pinned one — a version about four times the size (40,609 bytes, 692 lines, from 2026-09-30) whose questions adapted to the kind of file and to the change itself, covered notebook edits, let comment-only edits pass with a note, counted a search already done as the check, and added optional decision metrics. That version was never the pinned one, and the fingerprint in the old prompt did not match the pinned file, so its integrity check would have failed on the file this prompt installs. This version describes the pinned version exactly: 9,973 bytes, 207 lines, at commit 4eb71d92a39cab44ad40ac9d8a6a5ccb4029d6c2. What changed: the gates now ask for fixed sets of facts (four for edits and new files, three for destructive Bash, two for routine Bash), with no adaptation by file type or by change, no notebook coverage, no comment-only exception, no decision metrics, and no systematic full check of sensitive files. What did not change: the three-stage gate (deny, force facts, allow retry), the exempt-globs setting, the session off switch, the hook script, the per-session denial cap, and the A/B result of +2.25 quality points. 1. Download the SKILL.md file into a temporary folder. 2. Compare the file with the fingerprint from Skill Harbor's review of 2026-11-17: - SKILL.md: SHA-256 03bb5ce5916659b5d6eb1b3acc5f81950dcdcdb0dfe389fe04b11b67d91e9b45 If it does not match, the file changed since the review: do not install it yet — tell me what you found and let me decide. 3. Read the file and tell me in plain words: what the skill does when used, whether it connects to a service, whether it reads or sends any personal information or keys, and anything that seems unusual. 4. When I say yes, copy SKILL.md into my skills folder, in a folder named "gateguard". 5. Check the copy opens correctly (valid frontmatter, files in place) and tell me what was installed, where, and what is left for me to do (for example, switching the hook on through hooks.json, and setting GATEGUARD_EXEMPT_GLOBS for folders I want left alone). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Rules: only touch the temporary folder and the install target. If something seems off, stop and ask me.

🔒 Security, provenance and what this touches
🧭 What this touches
Based on the install prompt we scanned, this skill does not run commands, make third-party network calls, install packages, write files, or ask for API keys.
Install prompt analysis · 2026-10-13
🔒 Security and provenance
📌 Pinned version: commit 4eb71d9 · 2026-10-13
🧪 Install ran in our test sandbox · Nov 11, 2026
👁 Observed during the install test · Nov 11, 2026 : no network contact
🔎 Last checked · Nov 21, 2026 : no change detected in the files linked by this install prompt
Skill Harbor never reviews the code of a listed skill. These are dated facts about the install prompt and our install test, not a safety certification. Review a skill yourself before you install it. How we verify
Wanna be extra careful?

The install prompt below already includes the vetting steps: your agent follows the community checklist before installing anything with executable code. Want more?

How to check a skill before installing →

✓

Verification certificate

Dated facts about what was checked for this listing, and how. This is not a safety certification.

Pinned version
4eb71d9 · 2026-10-13
Sandbox install test
Install ran in our test sandbox · Nov 11, 2026
Creator identity
Not verified
Fingerprint watch
Last checked · Nov 21, 2026 : no change detected in the files linked by this install prompt
Install prompt analysis
2026-10-13 : No commands, network calls, packages, file writes or API keys detected.

Skill Harbor never reviews the code of a listed skill. These are dated facts about the install prompt and our install test, not a safety certification. Review a skill yourself before you install it. How we verify

?

Questions

How do I install a skill?

Every product page includes a copy-paste install prompt. Paste it into your Muse and it sets the skill up for you — no manual configuration.

Where does my money go?

Straight to the seller. Skill Harbor never processes payments: checkout happens on the seller’s own page, usually Stripe.

What does the ✓ next to a creator’s name mean?

It means we confirmed the identity of the person behind the listing. It says nothing about the code itself — always check a skill before installing it.