Harden cross-platform skill installer packaging

.gitignore

@@ -0,0 +1,3 @@
+__pycache__/
+*.py[cod]
+.DS_Store

AGENTS.md

@@ -0,0 +1,42 @@
+# Repository Guidance for Codex Agents
+
+## Project purpose
+
+`skills-red` is a cross-platform offensive-security skill library for Claude and Codex, forked from Claude Red. The repository ships portable `SKILL.md` files for authorized red-team, penetration-testing, bug-bounty, CTF, and security-research workflows.
+
+## Safety boundaries
+
+- Keep examples scoped to authorized testing and lab/CTF contexts.
+- Do not add real victim infrastructure, credentials, customer data, or destructive defaults.
+- Preserve responsible-disclosure and evidence-handling guidance when editing reporting or exploit content.
+
+## Skill format
+
+- Skills live at `Skills/<category>/<skill-name>/SKILL.md`.
+- The folder name must match the frontmatter `name` value.
+- Codex uses the frontmatter `name` and `description` for skill discovery; keep descriptions specific and trigger-rich.
+- Prefer concise, operator-useful instructions over generic background.
+- Use language tags on code blocks.
+
+## Codex packaging expectations
+
+- Codex installs skills as `$CODEX_HOME/skills/<skill-name>/SKILL.md` (default `$CODEX_HOME` is `~/.codex`).
+- `./install.sh --platform codex` flattens category folders into individual Codex skill directories.
+- `./install.sh --platform claude` preserves the Claude-compatible category tree.
+- Regenerate manifests after skill or metadata changes:
+
+```bash
+python3 tools/build_manifest.py
+```
+
+## Verification
+
+Before claiming completion for repository changes, run the narrowest relevant checks:
+
+```bash
+python3 tools/build_manifest.py
+./install.sh --platform codex --dry-run
+./install.sh --platform claude --dry-run
+```
+
+For shell changes, also run `bash -n install.sh`.

CHANGELOG.md

@@ -1,9 +1,20 @@
 # Changelog
 
-All notable changes to `claude-red` are documented here. The library follows a phased roadmap (see [README.md](README.md#roadmap)). Versions follow [Semantic Versioning](https://semver.org/) where breaking changes mean skill renames, removals, or category restructures.
+All notable changes to `skills-red` are documented here. The library follows a phased roadmap (see [README.md](README.md#roadmap)). Versions follow [Semantic Versioning](https://semver.org/) where breaking changes mean skill renames, removals, or category restructures.
 
 ## [Unreleased]
 
+### Added
+
+- Cross-platform installation paths in `install.sh` for Codex, Claude, and OpenCode, selected explicitly by prompt or `--platform`.
+- `codex-skills.json` and `opencode-skills.json` manifests generated alongside the Claude-compatible manifest.
+- Root `AGENTS.md` with cross-platform repository operating guidance.
+
+### Changed
+
+- README and contributor guidance now document Codex, Claude, and OpenCode as first-class installation targets with no default installer platform.
+- Legacy platform-specific skill headings now use platform-neutral `Operational Instructions`.
+
 ### Planned
 
 - Phase 1 — Internal AD/Windows split (16 skills)

CONTRIBUTING.md

@@ -1,11 +1,11 @@
-# Contributing to claude-red
+# Contributing to skills-red
 
 Thanks for contributing. This guide explains the skill format, the review process, and the conventions to keep the library coherent as it grows.
 
 ## Quick Rules
 
 1. **One skill, one surface.** Prefer focused skills (`offensive-kerberoasting`) over monolithic overviews (`offensive-active-directory`).
-2. **YAML frontmatter is required.** Skills without it won't load via the Claude Skills system.
+2. **YAML frontmatter is required.** Skills without it won't load reliably via Codex or Claude skill discovery.
 3. **Cite sources.** Every technique should be attributable. Link CVEs, advisories, original research.
 4. **No unauthorized targeting.** Don't include hardcoded victim domains, real customer data, or credentials.
 5. **Use code blocks with language tags.** It's how Claude (and humans) parse them best.
@@ -27,11 +27,11 @@ The folder name **must** match the `name:` field in the frontmatter.
 ```yaml
 ---
 name: offensive-<bug-class-or-domain>
-description: "One paragraph (50–500 words). State the surface, the techniques covered, and when to use this skill. Claude uses this for trigger matching — be specific about scenarios, tools, and sub-topics."
+description: "One paragraph (50–500 words). State the surface, the techniques covered, and when to use this skill. Codex uses this for trigger matching — be specific about scenarios, tools, and sub-topics."
 ---
 ```
 
-The `description` is what Claude matches against. Make it dense with relevant terms an operator would mention. Avoid marketing language.
+The `description` is what Codex and Claude use for trigger matching. Make it dense with relevant terms an operator would mention. Avoid marketing language.
 
 ### Body Structure (recommended)
 
@@ -89,8 +89,13 @@ The `description` is what Claude matches against. Make it dense with relevant te
 3. Write the frontmatter and body following the structure above.
 4. Update [`README.md`](README.md) — add the skill to the relevant category table.
 5. Update [`CHANGELOG.md`](CHANGELOG.md) under the next version.
-6. Update [`claude-skills.json`](claude-skills.json) if it exists (run `python tools/build_manifest.py` if available).
-7. Run any local lint:
+6. Update the generated manifests (`claude-skills.json`, `codex-skills.json`, and `opencode-skills.json`) by running `python3 tools/build_manifest.py`.
+7. Confirm the generated manifests are fresh:
+   ```bash
+   python3 tools/check_manifest_fresh.py
+   ```
+   Manifest install paths preserve `skills-red/<category>/<skill-name>` under each platform skill root; update `tools/platform_defaults.sh` rather than editing generated JSON by hand.
+8. Run any local lint:
    ```bash
    ./tools/check-skill.sh Skills/<category>/<skill-name>/SKILL.md
    ```
@@ -107,7 +112,7 @@ When a skill grows beyond one surface (e.g. `offensive-wifi` covering WPA2, WPA3
 
 1. Keep the original as a brief overview that points to the new focused skills
 2. Move detailed content into new per-surface skills
-3. Update README, CHANGELOG, and the manifest
+3. Update README, CHANGELOG, and the generated manifests
 
 The roadmap in README tracks current splits.
 
@@ -131,7 +136,7 @@ Expect one round of review. Maintainers may request edits before merging.
 - Tooling that has destructive defaults without warnings
 - Bypasses for vendor-mandated security telemetry without legitimate red team context
 - Content under non-MIT-compatible licenses
-- AI-generated skills without operator review (use Claude to draft, then verify and edit)
+- AI-generated skills without operator review (use Codex or Claude to draft, then verify and edit)
 
 ---
 

MINDMAP.md

@@ -1,4 +1,4 @@
-# claude-red — Library Mindmap
+# skills-red — Library Mindmap
 
 A visual map of every skill in the library, by category. Use it to navigate, to discover skills you didn't know existed, and to spot coverage gaps before an engagement.
 
@@ -10,7 +10,7 @@ References for completeness checking: [MITRE ATT&CK](https://attack.mitre.org/),
 
 ```mermaid
 mindmap
-  root((claude-red))
+  root((skills-red))
     Web
       sqli
       xss

README.md

@@ -1,16 +1,16 @@
-![claude-red banner](/assets/banner.png)
+![skills-red banner](/assets/banner.png)
 
 <div align="center">
 
-# claude-red
+# skills-red
 
-**Offensive security skills for Claude — drop-in `SKILL.md` files that turn Claude into a context-aware red team operator.**
+**Offensive security skills for Claude, Codex, and OpenCode — portable `SKILL.md` files that turn AI coding agents into context-aware red team operators.**
 
 [![License: MIT](https://img.shields.io/badge/license-MIT-blue.svg)](LICENSE)
 [![Skills](https://img.shields.io/badge/skills-58-red.svg)](#skill-index)
 [![Categories](https://img.shields.io/badge/categories-13-orange.svg)](#categories)
-[![Stars](https://img.shields.io/github/stars/SnailSploit/claude-red?style=social)](https://github.com/SnailSploit/claude-red)
-[![Forks](https://img.shields.io/github/forks/SnailSploit/claude-red?style=social)](https://github.com/SnailSploit/claude-red/network/members)
+[![Stars](https://img.shields.io/github/stars/trewwwsec/skills-red?style=social)](https://github.com/trewwwsec/skills-red)
+[![Forks](https://img.shields.io/github/forks/trewwwsec/skills-red?style=social)](https://github.com/trewwwsec/skills-red/network/members)
 
 Built by **[SnailSploit](https://snailsploit.com)** — GenAI Security Research.
 
@@ -22,6 +22,7 @@ Built by **[SnailSploit](https://snailsploit.com)** — GenAI Security Research.
 
 - [What is this](#what-is-this)
 - [Quickstart](#quickstart)
+- [Generated manifests](#generated-manifests)
 - [Categories](#categories)
 - [Skill Index](#skill-index)
   - [Web Application](#web-application)
@@ -46,25 +47,70 @@ Built by **[SnailSploit](https://snailsploit.com)** — GenAI Security Research.
 
 ## What is this
 
-`claude-red` is a curated library of offensive security skills for the [Claude Skills system](https://docs.claude.com). Each skill is a structured `SKILL.md` file that primes Claude with expert-level methodology for a specific attack surface — from SQLi to shellcode, EDR evasion to ADCS abuse.
+`skills-red` is a curated library of offensive security skills for Claude, Codex, and OpenCode skill systems. Each skill is a structured `SKILL.md` file that primes an AI coding agent with expert-level methodology for a specific attack surface — from SQLi to shellcode, EDR evasion to ADCS abuse.
 
-Drop a skill into your Claude environment and it behaves like a specialist: it knows the techniques, the tooling, the edge cases, and the escalation paths. Skills load on demand based on conversational triggers — you don't pay context for skills you aren't using.
+Install skills into `$CODEX_HOME/skills/skills-red` for Codex, `~/.claude/skills/skills-red` for Claude, or `~/.config/opencode/skills/skills-red` for OpenCode. The same `SKILL.md` files are portable across all supported platforms; the installer preserves the category tree under the `skills-red` namespace for each platform.
 
 **Use it for:** authorized red team engagements, bug bounty triage, security research, CTF preparation, training operators, and exploring attack surfaces methodically.
 
 ---
 
 ## Quickstart
 
-### Claude Skills System (recommended)
+### Codex Skills System
 
 ```bash
-# Clone into a directory Claude will scan
-git clone https://github.com/SnailSploit/claude-red ~/.claude/skills/claude-red
+# Clone this repo and install all skills into ~/.codex/skills/skills-red/<category>/<skill-name>
+git clone https://github.com/trewwwsec/skills-red
+cd skills-red
+./install.sh --platform codex
 
 # Or install only one category
-git clone --filter=blob:none --sparse https://github.com/SnailSploit/claude-red
-cd claude-red && git sparse-checkout set Skills/web Skills/active-directory
+./install.sh --platform codex --category web
+```
+
+Restart Codex after installation so the skill metadata is picked up. Current Codex releases recursively discover `SKILL.md` files under `$CODEX_HOME/skills`, so the installer keeps skills namespaced as `$CODEX_HOME/skills/skills-red/<category>/<skill-name>/SKILL.md`.
+
+### Manual Codex install
+
+```bash
+mkdir -p ~/.codex/skills/skills-red/web
+cp -R Skills/web/offensive-sqli ~/.codex/skills/skills-red/web/offensive-sqli
+```
+
+### OpenCode Agent Skills
+
+```bash
+# Install all skills into ~/.config/opencode/skills/skills-red/<category>/<skill-name>
+git clone https://github.com/trewwwsec/skills-red
+cd skills-red
+./install.sh --platform opencode
+
+# Or install only one category
+./install.sh --platform opencode --category web
+```
+
+Start a new OpenCode session after installation so the `skill` tool refreshes its available skills. OpenCode's public docs show the simple one-folder layout, but current OpenCode releases recursively scan `SKILL.md` files under configured skill roots (`{skill,skills}/**/SKILL.md` in the upstream loader). The installer relies on that current recursive discovery behavior to keep skills namespaced as `~/.config/opencode/skills/skills-red/<category>/<skill-name>/SKILL.md`.
+
+### Manual OpenCode install
+
+```bash
+mkdir -p ~/.config/opencode/skills/skills-red/web
+cp -R Skills/web/offensive-sqli ~/.config/opencode/skills/skills-red/web/offensive-sqli
+```
+
+### Claude Skills System
+
+```bash
+# Upstream Claude Red install into a directory Claude will scan
+git clone https://github.com/SnailSploit/claude-red ~/.claude/skills/claude-red
+```
+
+```bash
+# Or install skills-red while preserving Claude's category tree
+git clone https://github.com/trewwwsec/skills-red
+cd skills-red
+./install.sh --platform claude
 ```
 
 Claude will auto-load matching skills based on conversational triggers (e.g. mentioning SQLi loads `offensive-sqli`).
@@ -79,20 +125,48 @@ cat Skills/web/offensive-sqli/SKILL.md | claude --system-file -
 cat Skills/active-directory/**/SKILL.md | claude --system-file -
 ```
 
-### Claude.ai (Manual)
-
-Paste the contents of a `SKILL.md` into a Project's system prompt or prepend to your conversation.
-
 ### Install Script
 
 ```bash
-./install.sh                           # interactive
-./install.sh --target ~/.claude/skills # explicit target
+./install.sh                           # prompt for platform, then target
+./install.sh --platform codex          # Codex install
+./install.sh --platform claude         # Claude-compatible install
+./install.sh --platform opencode       # OpenCode agent-skill install
+./install.sh --target ~/.codex/skills/skills-red  # prompt for platform, explicit target
+./install.sh --platform opencode --target ~/.config/opencode/skills/skills-red  # explicit OpenCode target
 ./install.sh --category web            # one category
+./install.sh --dry-run                 # prompt for platform and preview copy plan
 ```
 
 ---
 
+## Generated manifests
+
+The root `claude-skills.json`, `codex-skills.json`, and `opencode-skills.json`
+files are generated distribution indexes, not hand-authored skill sources. They
+summarize the `Skills/<category>/<skill-name>/SKILL.md` tree for platform
+tooling, marketplace/index consumers, and release review while keeping each
+`SKILL.md` file as the source of truth.
+
+Manifest `install_path` values are install-location metadata for indexes and review. They intentionally include the `skills-red/<category>/<skill-name>` namespace so generated indexes match each platform's target layout after the installer prompts for, or receives, an explicit platform. For Codex and OpenCode this depends on current recursive skill discovery; if either platform drops recursive scanning, update `tools/platform_defaults.sh`, `install.sh`, and regenerated manifests together. Runtime installation behavior, including platform selection, `--target`, `CODEX_HOME`, and `OPENCODE_CONFIG_HOME` overrides, is owned by [`install.sh`](install.sh).
+
+Regenerate them after any skill metadata, category, or install-path change:
+
+```bash
+python3 tools/build_manifest.py
+python3 tools/check_manifest_fresh.py
+```
+
+The freshness check also validates manifest `install_path` values against the
+shared defaults in `tools/platform_defaults.sh`, which are consumed by both the
+installer and manifest tooling.
+
+Do not edit the root manifest JSON files by hand; update the relevant
+`SKILL.md` frontmatter or `tools/build_manifest.py`, then regenerate and commit
+the resulting manifest diff.
+
+---
+
 ## Categories
 
 | Category | Skills | Focus |
@@ -308,7 +382,7 @@ Contributions welcome. See [CONTRIBUTING.md](CONTRIBUTING.md) for the skill temp
 
 <div align="center">
 
-> *"Give Claude the right skill and it stops being a chatbot. It becomes an operator."*
+> *"Give the agent the right skill and it stops being a chatbot. It becomes an operator."*
 
 </div>
 
@@ -320,7 +394,7 @@ Contributions welcome. See [CONTRIBUTING.md](CONTRIBUTING.md) for the skill temp
 
 This project's full writeup, methodology, and related research lives at:
 
-**[https://snailsploit.com/claude-red](https://snailsploit.com/claude-red)**
+**Upstream:** [https://snailsploit.com/claude-red](https://snailsploit.com/claude-red)
 
 Created by **Kai Aizen** — independent offensive security researcher.
 

SECURITY.md

@@ -1,6 +1,6 @@
 # Security Policy
 
-`claude-red` is an offensive security tooling library. Its content describes attack methodologies for use by authorized red team operators, penetration testers, and security researchers.
+`skills-red` is an offensive security tooling library. Its content describes attack methodologies for use by authorized red team operators, penetration testers, and security researchers.
 
 ## Intended Use
 
@@ -13,7 +13,7 @@ These skills are intended for:
 
 These skills are **not** intended for unauthorized access to systems you do not own or do not have explicit, written permission to test. Misuse may violate computer-misuse laws in your jurisdiction (CFAA in the US, Computer Misuse Act in the UK, equivalent statutes elsewhere).
 
-## Reporting a Vulnerability in claude-red Itself
+## Reporting a Vulnerability in skills-red Itself
 
 If you discover a security issue in this repository — for example a malicious payload accidentally committed, a credential leaked in an example, a typosquat-prone install path, or an unsafe shell command in `install.sh` — please report it privately rather than opening a public issue.
 
@@ -30,7 +30,7 @@ We aim to acknowledge reports within 72 hours and resolve confirmed issues withi
 
 ## Reporting a Vulnerability Found Using This Library
 
-If you discover a vulnerability in a third-party product or service while using `claude-red`'s methodologies, follow that vendor's responsible disclosure process. The [`offensive-reporting`](Skills/utility/offensive-reporting/SKILL.md) skill includes guidance on responsible disclosure, evidence handling, and report writing.
+If you discover a vulnerability in a third-party product or service while using `skills-red`'s methodologies, follow that vendor's responsible disclosure process. The [`offensive-reporting`](Skills/utility/offensive-reporting/SKILL.md) skill includes guidance on responsible disclosure, evidence handling, and report writing.
 
 If the vendor has no published security contact:
 
@@ -47,7 +47,7 @@ This repository is signed by SnailSploit. Verify commit signatures with:
 git log --show-signature
 ```
 
-If you receive a `claude-red` archive from a third party (mirror, pastebin, package manager), verify it against the upstream repository before using.
+If you receive a `skills-red` archive from a third party (mirror, pastebin, package manager), verify it against the upstream repository before using.
 
 ## Scope
 

Skills/ai/offensive-ai-security/SKILL.md

@@ -1,3 +1,8 @@
+---
+name: offensive-ai-security
+description: "AI/LLM security offensive checklist: prompt injection, jailbreaking, model extraction, training data poisoning, adversarial inputs, LLM-assisted attack automation, and AI system reconnaissance. Use when assessing AI/ML systems, red-teaming LLMs, or researching AI attack vectors."
+---
+
 # SKILL: AI Pentest
 
 ## Metadata
@@ -12,7 +17,7 @@ AI/LLM security offensive checklist: prompt injection, jailbreaking, model extra
 Use this skill when the conversation involves any of:
 `AI security, LLM security, prompt injection, jailbreak, model extraction, training data poisoning, adversarial input, AI red team, ML security, RAG poisoning, AI attack`
 
-## Instructions for Claude
+## Operational Instructions
 
 When this skill is active:
 1. Load and apply the full methodology below as your operational checklist