# Thiscodex > Use when setting up an OpenAI Codex CLI agent as a persistent Discord bot with Claude-Code-equivalent persona/vault discipline, when porting Claude Code skills or rules to Codex, or when wiring multi-agent conventions (cross-bot addressing, meeting threads, SessionStart roster, progressive-disclosure rules) so Claude Code and Codex agents share one Discord + Obsidian-vault workspace. Covers headl… - **Type:** Skill - **Install:** `agentstack add skill-treylom-thiscodex-thiscodex` - **Verified:** Yes — security-reviewed for prompt injection and unsafe behavior - **Seller:** [treylom](https://agentstack.voostack.com/s/treylom) - **Installs:** 0 - **Category:** [AI & ML](https://agentstack.voostack.com/c/ai-and-ml) - **Latest version:** 0.1.0 - **License:** MIT - **Upstream author:** [treylom](https://github.com/treylom) - **Source:** https://github.com/treylom/ThisCodex/tree/master/skills/thiscodex ## Install ```sh agentstack add skill-treylom-thiscodex-thiscodex ``` Requires the [AgentStack CLI](https://agentstack.voostack.com/docs/cli). Works with Claude Code, Cursor, and any MCP-compatible agent. ## About # ThisCodex — make Codex behave like a Claude Code Discord bot ## Install — read this before doing anything **Default install = interactive guided onboarding.** Run: ``` thiscodex init ``` (or `npx github:treylom/ThisCodex init`). For an AI agent handed this repo, the active session/TUI is the user communication channel: **run guided `thiscodex init` and relay every question to the user**, feeding confirmed answers back — one question at a time. Do **not** auto-run a non-interactive install, and do **not** fabricate an answers file to skip the questions. Copying this skill folder is *placement*; **placement is not guided onboarding** — do **not** report "copied = installed" or claim the bot is installed from placement alone. Use `--non-interactive` / `--yes` / `--answers` only when the user explicitly asks for CI or automation. If the installer stops on a missing required decision, it prints an interactive-recovery hint — relay that to the user and let them choose; never self-answer to push past it. Invocable skill. Install paths: - **user-tier**: copy this folder to `~/.agents/skills/thiscodex/` (codex 4-layer scan picks it up; deep docs are referenced by URL below so a loose copy still works). - **marketplace/plugin** — ⚠️ **incomplete on codex 0.130, not a usable install today; do not use as the entry path** (verified code-review-bot, 2026-05-18, raw): the repo ships `.codex-plugin/plugin.json`, but `codex plugin marketplace add treylom/ThisCodex` **fails** — `marketplace root does not contain a supported manifest`; codex requires a `.agents/plugins/marketplace.json` the repo does not yet ship. There is **no `codex plugin install`** subcommand (only `marketplace add/upgrade/remove`); cwd auto-load of `.codex-plugin/plugin.json` does not work; `tool_search thiscodex` = 0. The Codex App `/plugins` GUI auto-recognition is unverified from CLI (separate GUI check needed). **Until future `.agents/plugins/marketplace.json` packaging lands, use the user-tier loose-skill copy above or `thiscodex init` (guided) — not the plugin path.** Deep reference lives in the repo (load only when a step needs it — progressive disclosure). ## When to use - "Make my Codex CLI a Discord bot like the Claude Code ones" - "Port these Claude Code skills/rules to Codex" - "Set up multi-agent (Claude + Codex) in one Discord + vault" ## Setup procedure (in order) 1. **Prereqs**: `codex` CLI, `tmux`, Python 3 + `websockets`, the Claude Code Discord plugin (reused as a codex MCP server), `gh auth login`. Platforms: macOS / Linux / WSL2. 2. **`~/.codex/config.toml`** — auto-load persona/rules + wire the Discord MCP: ```toml project_doc_fallback_filenames = ["SOUL.md", "AGENTS.md"] project_doc_max_bytes = 65536 [mcp_servers.discord] command = "bun" args = ["run", "--cwd", "", "start"] [mcp_servers.discord.env] DISCORD_STATE_DIR = "~/.claude/channels/discord-" ``` 3. **Bot working directory**: put `SOUL.md` (persona) + `AGENTS.md` (rules) there. `AGENTS.md` carries the static Discord-reply rule and points at `rules/INDEX.md` only (not inline rules). 4. **Bridge + launcher**: use the shipped **`../../scripts/launch.sh`** (hardened 2-window tmux launcher: `infra` runs app-server + `bot.py` bridge; `codex` resumes the SAME bridge thread). It enforces the invariants — command-as-window-process (never `send-keys` into a bare shell), and the codex window **always `codex resume --remote`, never a bare fresh `codex --remote`**. Set `BOT_WD`, `SESSION`, `LAUNCH_CMD`. Do not hand-roll this (see Troubleshooting for why). 5. **YOLO**: `bot.py` sends `sandbox:"danger-full-access"` + `approvalPolicy:"never"` on **both** `thread/start` AND `thread/resume` (resume silently degrades otherwise — the nastiest bug). 6. **Skills/rules portability**: your own skills → `~/.agents/skills//SKILL.md`. Plugin/framework skills (e.g. superpowers) → install via the framework's own codex path, never hand-symlink. 7. **Multi-agent conventions**: cross-bot `` addressing, meeting = dedicated thread, SessionStart roster injection — all from one `bot-roster.yaml`. 8. **Rules without context bloat**: ship a `rules/INDEX.md` router + on-demand topical files; `CLAUDE.md`/`AGENTS.md` point only at the router. ## Verify (don't claim done without this) - `codex features list | grep skill_` on both targets - one clean `dangerFullAccess` turn → tool list has `web.run, exec_command, image_gen` (no browser/computer tool — `computer_use` is parked upstream) - bridge log shows `[READY]` + a `thread/resume` with `{"type":"dangerFullAccess"}` ## Smoke test (expected trigger) > Prompt: "set up codex as a discord bot like claude code" → this skill should activate and produce the 8-step procedure above with the verify gate. If it does not trigger on that phrasing, the `description` frontmatter needs widening. ## Troubleshooting **Symptom**: Discord messages reach the bot (the `infra` window logs turns, replies get sent) but the **codex TUI window shows nothing / a fresh empty prompt** — "infra catches it, the codex TUI doesn't". **Cause**: the codex window ran `codex --remote ws://…` (a FRESH session/thread) instead of `codex resume "$(cat .codex-thread-id)" --remote ws://…`. `bot.py` drives the bridge thread; the TUI is on a different empty thread, so it never shows the bridge's turns. Common when the launcher is hand-rolled, or a stale launcher definition is used on manual recovery. **Fix**: the codex window MUST `codex resume --remote` the SAME thread `bot.py` uses (`.codex-thread-id`) — never bare `codex --remote`. `../../scripts/launch.sh` invariant 2 enforces this; use it instead of hand-rolling. Fix a live session without a full restart (bot.py/infra untouched): ``` tmux respawn-window -k -t :codex -c \ "codex resume $(cat /.codex-thread-id) --remote ws://127.0.0.1:4222" ``` **Symptom**: multi-bot collaboration channel — bots (Claude *and* this Codex bot) can't read each other's messages even though every bot id is in `access.json`. **Cause**: the shared official discord plugin (`…/external_plugins/discord/server.ts`, ≈L806) does `if (msg.author.bot) return` **before** `gate()` — every bot-authored message is dropped pre-allowlist. This Codex bot reuses that same plugin, so it is hit identically. Root cause is in the external plugin, **not** ThisCodex. **Fix**: replace the blanket `msg.author.bot` drop with the 3-guard (self-loop / webhook / bot-DM block, else fall through to `gate()`). Full recipe: ThisCode `docs/08-debug-노하우.md` **J-2**. Permanent re-apply layer is **built** — ThisCode `scripts/patch-discord-bot-drop.sh` (idempotent, fail-open, .bak, exact-match-only), wired into `/thiscode:self-update pull` + opt-in SessionStart. Reuse it on the Codex side too (same external plugin). Decision record: `docs/2026-05-18-repo-handoff-interactive-default-design.md` §10.1. ## Subcommands When you invoke `/thiscodex`, the skill routes by intent. Common subcommands: > **Implementation note**: only `init`, `doctor`, and `smoke` are real CLI > commands (`npx github:treylom/ThisCodex ` / `bin/thiscodex.mjs`). The > other rows (`port-skills`, `multi-agent`, `run`, `logs`, `features`, > `troubleshoot`) are **AI-guided intents** — the assistant reading this skill > performs them by following the linked reference docs (e.g. `run` = > `scripts/launch.sh` per docs/bot-launch-pattern.md). They are not standalone > shell executables; `thiscodex run` typed directly in a terminal will fail by design. | When to use | Call | |---|---| | Start guided setup | `/thiscodex init` — launches interactive onboarding for repo, workspace, BOT_WD, Discord MCP, Codex hooks, tmux aliases. **Start here** for a fresh Codex bot. | | Verify setup readiness | `/thiscodex doctor` — runs full diagnostic (paths, Discord MCP, config.toml, hook trust hashes, tmux, Python websockets). Shows what's missing and how to fix it. | | Port Claude Code skills | `/thiscodex port-skills` — adds a Claude Code skill to your Codex skills directory with path adjustments (e.g., `CLAUDE.md` → `AGENTS.md`). | | Set up multi-agent conventions | `/thiscodex multi-agent` — generates `bot-roster.yaml`, wires cross-bot addressing, meeting thread rules, and SessionStart roster injection for Claude Code + Codex coexistence. | | Launch the bot (tmux) | `/thiscodex run` — starts the infra (app-server + bot.py bridge) and codex TUI windows inside tmux session. Assumes setup is done. | | Tail bot logs | `/thiscodex logs` — shows live `bot.py` and app-server logs for debugging turns, tool calls, or Discord events. | | Check feature availability | `/thiscodex features` — lists what's working (Discord bot, multi-client same-thread, YOLO, image_gen, web.run) and what's parked (computer_use/browser_use). | | Generate troubleshooting guide | `/thiscodex troubleshoot ` — provides step-by-step fixes. Example: `/thiscodex troubleshoot "codex tui shows nothing"` for the common "TUI on wrong thread" issue. | ## Reference map (load on demand — GitHub URLs, robust for loose-copy or marketplace install) | Need | Source | |---|---| | Architecture, protocol, evidence | [README.md](https://github.com/treylom/ThisCodex/blob/master/README.md) · [README.ko.md](https://github.com/treylom/ThisCodex/blob/master/README.ko.md) | | CC↔Codex skill porting, superpowers path | [docs/skill-portability.md](https://github.com/treylom/ThisCodex/blob/master/docs/skill-portability.md) | | Progressive-disclosure rules convention | [docs/rules-system.md](https://github.com/treylom/ThisCodex/blob/master/docs/rules-system.md) | | Session↔channel UX (borrowed) | [ThisCode/docs/connector-session-ux.md](https://github.com/treylom/ThisCode/blob/main/docs/connector-session-ux.md) | Companion runtime (Claude Code side): [ThisCode](https://github.com/treylom/ThisCode). This repo = the Codex side + cross-runtime conventions. ## Source & license This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code. - **Author:** [treylom](https://github.com/treylom) - **Source:** [treylom/ThisCodex](https://github.com/treylom/ThisCodex) - **License:** MIT Install and usage instructions live in the source repository linked above. ## Pricing - **Free** — Free ## Security capabilities Automated source analysis of v0.1.0 — what this tool can access: - **Network access:** no - **Filesystem access:** no - **Shell / process execution:** no - **Environment & secrets:** yes - **Dynamic code execution:** no *"Yes" means the capability is present in the source — more access means more to trust, not that it is unsafe.* ## Versions - **0.1.0** — security scan: passed — Imported from the upstream source. ## Links - Listing page: https://agentstack.voostack.com/l/skill-treylom-thiscodex-thiscodex - Seller: https://agentstack.voostack.com/s/treylom - Browse the marketplace: https://agentstack.voostack.com/browse --- Listed on AgentStack — the marketplace for AI agent skills and MCP servers. Every listing is security-reviewed. Creators keep 70%.