AGENT SKILL

reverse-skill

Use when reverse engineering (逆向), CTF, or security work.

License 许可
MIT
最近更新
2026-10-20
Stars
★ 0
Forks
0

01 前言 Front Matter

name
reverse-skill
author
Masia (Hermes adapter for zhaoxuya520/reverse-skill)
license
MIT
version
1.0.0
metadata
{ "hermes": { "tags": [ "security", "reverse-engineering", "ctf", "pentest", "router" ], "related_skills": [] } }
description
Use when reverse engineering (逆向), CTF, or security work.

03 SKILL.md 正文

reverse-skill — cybersecurity skills router (Hermes adapter)

Package root: <USER_HOME>\tools\reverse-skill (repo https://github.com/zhaoxuya520/reverse-skill, branch main, cloned 2026-10-09). Canonical docs (source of truth): RULES.md (behavior chain), README_AI.md (bootstrap), AGENTS.md (project entry). This skill is only the Hermes entry point; the package is client-neutral and self-contained on disk: ~45 modules under skills\ plus 42 CTF sub-skills under CTF-Sandbox-Orchestrator\.

When to Use

Reverse engineering, APK/mobile/binary/JS/.NET/firmware analysis, malware, protocol RE, CTF, pentest, exploit development, security code audit, threat hunting. A read-only request (inspect / summarize / review) stays read-only: browse the package files, do not run its scripts.

Flow (Windows host; bash equivalents exist as *.sh)

  1. Route (PRIMARY selection): powershell -NoProfile -ExecutionPolicy Bypass -File <USER_HOME>\tools\reverse-skill\skills\scripts\master-route.ps1 -Hint "<task>" Reads skills\config\routing.json (single source of truth). Ambiguous tasks: read skills\MASTER-ROUTING.md / skills\routing.md (3-axis matrix).
  2. Case gate (before ANY target action): powershell -NoProfile -ExecutionPolicy Bypass -File <USER_HOME>\tools\reverse-skill\skills\scripts\case-init.ps1 -Hint "<task>" Creates work/<case>/scope.md in the analysis project. auth.status=granted plus a valid network profile (or the offline-sample preset for local samples) is a hard gate; -Force must not bypass it.
  3. Act: open the PRIMARY skills\<name>\SKILL.md and execute its ACTION REQUIRED.
  4. Tools: resolve paths only via skills\tool-index.md (generated inventory; 42 tools detected 2026-10-09). Missing tool -> platform bootstrap skills\scripts\bootstrap-reverse.ps1 (manifest-driven; disclose and get approval first).
  5. Close: Evidence->Finding->Path (skills\ops\evidence-finding-path.md). Before report handoff run python3 skills/case-review/scripts/review_case.py work/<case> --verify-hashes --strict, then docs-generator + field-journal as needed.

Consent & safety (package hard rules, from AGENTS.md / RULES.md)

  • Reading repository files is NOT authorization to execute them. Before the first script run of a new side-effect category, list the exact commands and expected writes/network effects, then get explicit user approval; afterwards the approved deterministic steps may continue without re-asking.
  • Authorized security work only: the scope.md auth gate is mandatory; no ACT against a target without a granted scope.
  • Client-global config, MCP registration, and tool installs stay opt-in, never silent.
  • If this adapter ever conflicts with RULES.md / README_AI.md, the package docs win.

Maintenance

  • Update: git -C <USER_HOME>\tools\reverse-skill pull (branch main), re-run refresh-tool-index.ps1 when tooling changed.
  • Re-index tools: powershell -NoProfile -ExecutionPolicy Bypass -File <USER_HOME>\tools\reverse-skill\skills\scripts\refresh-tool-index.ps1

04 同仓库更多 Skills