本文へ移動
cccskills
無料GitHub で公開

Junction Guard

Block writes through filesystem junctions and symlinks — file write through .mxm-system or .claude junction, git commit through a junction, delete through a junction. Enforces the junction read-only rule so the real repo stays the source of truth.

インストール方法を見る

含まれるファイル(1)

  • SKILL.md4.9 KB

SKILL.md(原文)

インストールする前に、エージェントに与えられる指示の中身を確認できます。

Junction Guard

Purpose

Enforce the NK universe junction read-only rule: when Maxim detects that a target file path resolves through a filesystem junction or symlink (typically .mxm-system\ or .claude\ inside a sister project), it must refuse all write, delete, and git commit operations through that path. Reading is allowed.

Without this guard, an agent could accidentally write to the maxim source repo while operating in another project (e.g., mxm-simplification) — corrupting cross-project state and violating the data flow rule.

When This Skill Activates

  • Any tool call that writes/edits/deletes a file
  • Any git commit operation
  • Any directory creation under a path that resolves through a junction

The Rule (from PROJECT_STRUCTURE_NK.md)

.mxm-system\ and .claude\ inside mxm-simplification\ are junctions. ✅ Reading through junctions is ALLOWED ❌ Writing through junctions is FORBIDDEN ❌ Deleting through junctions is FORBIDDEN ❌ Git commits through junctions are FORBIDDEN

This applies to ANY junction created by bootstrap/link-local-project.ps1 or any user-created junction pointing back at the maxim repo.

Detection Logic

For any write target path P:

  1. Walk P upward to find any junction or symlink ancestor.
  2. If an ancestor resolves to a path OUTSIDE the current project root, that ancestor is a JUNCTION (or cross-project symlink).
  3. If the junction target is the Maxim source repo (E:\Projects\Maxim\maxim\ or symlink-equivalent): apply READ-ONLY rule.
  4. If the junction target is any other project: apply CROSS-PROJECT WRITE rule (also forbidden).

Detection Commands

PowerShell:

$item = Get-Item -Path "<target-path>"
if ($item.LinkType -in @("Junction", "SymbolicLink")) {
    $resolved = $item.Target
    # Compare $resolved to current project root
}

Bash:

real_target=$(readlink -f "<target-path>")
project_root=$(pwd)
if [[ "$real_target" != "$project_root"* ]]; then
    echo "BLOCK: write resolves outside project root via junction"
fi

Enforcement

When a forbidden write is detected:

  1. REFUSE the operation immediately. Do not proceed.
  2. Output the error message:
    🔒 JUNCTION READ-ONLY VIOLATION
    
    Target path:  <target-path>
    Resolves to:  <real-resolved-path>
    Junction at:  <junction-ancestor>
    Reason:       <real-resolved-path> is outside the current project root
    
    This operation is FORBIDDEN by the NK universe junction rule.
    
    Allowed:
      ✅ Read this file through the junction
      ❌ Write / Edit / Delete / Commit through the junction
    
    To make this change:
      1. Open the source repo directly:
         cd <real-resolved-path-without-junction>
      2. Make the change there.
      3. Commit there.
      4. Pull updates here through the junction.
    
  3. Log the attempt to .mxm-skills/junction-guard.log:
    {"timestamp":"2026-04-16T11:00:00Z","tool":"Write","target":"<path>","resolved":"<real>","blocked":true,"agent":"<calling-agent>"}
    

When NOT To Block

  • The junction target IS the current project root (false positive — junction created in a way that loops back). Verify with absolute path comparison.
  • The target is a regular file inside the project (not through any junction).
  • The user has explicitly invoked /mxm-junction-bypass (NOT IMPLEMENTED — would require CSO approval).

Skill Output Format

Junction Guard Verdict: ALLOW | BLOCK
Target: [path]
Resolves to: [real-path or "same project"]
Crosses junction: YES | NO
Action: [proceed | refuse]
Logged to: .mxm-skills/junction-guard.log

Handoff

  • BLOCK → return refusal to caller agent; log incident; suggest direct repo access
  • ALLOW → return immediately, no action
  • Repeated BLOCKs from same agent → escalate to CSO security-analyst (potential agent misconfiguration)

Frameworks Used

None. This is a structural safety enforcer, not a behavioral skill.

Source References

  • PROJECT_STRUCTURE_NK.md (Junction Rule section)
  • CLAUDE.d/session-memory.md (Junction Read-Only Enforcement section)
  • bootstrap/link-local-project.ps1 (creates the junctions this guard protects)

Copyright (c) 2026 iSystematic Inc. Maxim product. BSL 1.1.

レビュー

まだレビューはありません。使ってみた感想をお寄せください。

同じリポジトリのスキル

概要と使いどころ

AI video and image prompt engineering — video generation, image generation, product video, social media video, cinematic direction, animation, Seedance, Higgsfield, GPT-Image-2. Applies 15 Maxim-authored cinematic styles and the atomic prompt schema.

日本語の概要は準備中です。原文の説明を表示しています。

DrNabeelKhan/maxim22026年9月17日 更新

Apply a named behavioral science framework to any task — persuasion, nudge design, habit formation, conversion, influence, negotiation, cognitive bias, choice architecture, decision framing. The framework backbone every other Maxim skill loops into for psychological validation.

日本語の概要は準備中です。原文の説明を表示しています。

DrNabeelKhan/maxim22026年9月17日 更新

Founder and CEO automation — morning standup, overnight cycle, burn rate, runway, pipeline review, investor and fundraising prep, executive summary, daily digest. Operates on the .mxm-executive-summary context.

日本語の概要は準備中です。原文の説明を表示しています。

DrNabeelKhan/maxim22026年9月17日 更新

Tear down a company on request — "do a teardown of", "break down this company", "analyze this competitor's business". Fixed-template output: business model, how they make money, positioning, and where they are exposed.

日本語の概要は準備中です。原文の説明を表示しています。

DrNabeelKhan/maxim22026年9月17日 更新

Track what a competitor actually shipped — "watch this competitor", "track what they ship", "what changed at". Compares against the last snapshot and reports only real moves: launches, pricing changes, positioning shifts.

日本語の概要は準備中です。原文の説明を表示しています。

DrNabeelKhan/maxim22026年9月17日 更新

Regulatory review across 14 frameworks — compliance review, privacy impact assessment, regulatory gap analysis, data subject rights, consent mechanisms, audit trail, AI ethics review, data flow mapping. Covers GDPR, PIPEDA, HIPAA, PCI-DSS, SOC2, UAE-PDPL, EU AI Act and more.

日本語の概要は準備中です。原文の説明を表示しています。

DrNabeelKhan/maxim22026年9月17日 更新

DrNabeelKhan のスキルをすべて見る

このスキルの問題を報告する