cue-ai 0.4.1 → 0.6.0

package/resources/mcps/configs/claude.sanitized.json

@@ -13,6 +13,61 @@
     },
     "cue-tty-watch": {
       "command": "/home/deadpool/Documents/cue/resources/mcps/cue-tty-watch/bin/cue-tty-watch"
+    },
+    "agentshield": {
+      "command": "/home/deadpool/Documents/cue/resources/mcps/agentshield/bin/agentshield"
+    },
+    "x-api": {
+      "command": "npm",
+      "args": [
+        "start"
+      ],
+      "cwd": "${HOME}/.x-api-mcp",
+      "env": {
+        "X_API_KEY": "${X_API_KEY}",
+        "X_API_SECRET": "${X_API_SECRET}",
+        "X_BEARER_TOKEN": "${X_BEARER_TOKEN}",
+        "X_ACCESS_TOKEN": "${X_ACCESS_TOKEN}",
+        "X_ACCESS_TOKEN_SECRET": "${X_ACCESS_TOKEN_SECRET}"
+      }
+    },
+    "trendradar": {
+      "command": "uvx",
+      "args": [
+        "--from",
+        "git+https://github.com/sansan0/TrendRadar.git",
+        "trendradar-mcp"
+      ],
+      "env": {}
+    },
+    "mapi-devdocs": {
+      "type": "http",
+      "url": "https://merchantapi.googleapis.com/devdocs/mcp/"
+    },
+    "mitre-attack": {
+      "command": "/home/deadpool/.local/bin/mitre-attack-mcp",
+      "args": [],
+      "env": {}
+    },
+    "cve-search": {
+      "command": "uv",
+      "args": [
+        "--directory",
+        "/home/deadpool/.config/cue/cache/cve-search_mcp",
+        "run",
+        "main.py"
+      ],
+      "env": {}
+    },
+    "shodan": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "@burtthecoder/mcp-shodan"
+      ],
+      "env": {
+        "SHODAN_API_KEY": "${SHODAN_API_KEY}"
+      }
     }
   },
   "source": "claude",

package/resources/mcps/configs/claude_runtime.sanitized.json

@@ -113,8 +113,69 @@
     },
     "cue-tty-watch": {
       "command": "/home/deadpool/Documents/cue/resources/mcps/cue-tty-watch/bin/cue-tty-watch"
+    },
+    "x-api": {
+      "command": "npm",
+      "args": [
+        "start"
+      ],
+      "cwd": "${HOME}/.x-api-mcp",
+      "env": {
+        "X_API_KEY": "${X_API_KEY}",
+        "X_API_SECRET": "${X_API_SECRET}",
+        "X_BEARER_TOKEN": "${X_BEARER_TOKEN}",
+        "X_ACCESS_TOKEN": "${X_ACCESS_TOKEN}",
+        "X_ACCESS_TOKEN_SECRET": "${X_ACCESS_TOKEN_SECRET}"
+      }
+    },
+    "reddit": {
+      "command": "npx",
+      "args": [
+        "reddit-mcp-server"
+      ],
+      "env": {
+        "REDDIT_CLIENT_ID": "${REDDIT_CLIENT_ID}",
+        "REDDIT_CLIENT_SECRET": "${REDDIT_CLIENT_SECRET}",
+        "REDDIT_USERNAME": "${REDDIT_USERNAME}",
+        "REDDIT_PASSWORD": "${REDDIT_PASSWORD}",
+        "REDDIT_SAFE_MODE": "standard",
+        "REDDIT_AUTH_MODE": "auto"
+      }
+    },
+    "google-ads-mcp": {
+      "command": "pipx",
+      "args": [
+        "run",
+        "--spec",
+        "git+https://github.com/googleads/google-ads-mcp.git",
+        "google-ads-mcp"
+      ],
+      "env": {
+        "GOOGLE_APPLICATION_CREDENTIALS": "${GOOGLE_APPLICATION_CREDENTIALS}",
+        "GOOGLE_PROJECT_ID": "${GOOGLE_PROJECT_ID}",
+        "GOOGLE_ADS_DEVELOPER_TOKEN": "${GOOGLE_ADS_DEVELOPER_TOKEN}"
+      }
+    },
+    "meta-ads": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "meta-ads-mcp@latest"
+      ],
+      "env": {
+        "META_ACCESS_TOKEN": "${META_ACCESS_TOKEN}",
+        "META_AD_ACCOUNT_ID": "${META_AD_ACCOUNT_ID}"
+      }
+    },
+    "mapi-devdocs": {
+      "command": "npx",
+      "args": [
+        "-y",
+        "mcp-remote",
+        "https://merchantapi.googleapis.com/devdocs/mcp"
+      ]
     }
   },
   "source": "claude_runtime",
   "source_path": "~/.claude.json"
-}
+}

package/resources/skills/README.md

@@ -1,135 +1,92 @@
-```
-   ┌───────────────────────────────────────────────────────────────┐
-   │                                                               │
-   │     r e c o d e e e  /  s k i l l s                           │
-   │     ─────────────────────────────────                         │
-   │     167 skills · 20 categories · one source of truth          │
-   │                                                               │
-   └───────────────────────────────────────────────────────────────┘
-```
-
-> **My laptop's current skill set.** This is the live snapshot of every agent
-> skill installed on my machine — the same folders that Claude Code and Codex
-> read at runtime. The repo is the source; my `~/.claude/skills` and
-> `~/.codex/skills` are just symlinks back here.
-
----
-
-## Why this exists
+# cue/skills — Skill Library
 
-I run a lot of agents. Skills drift fast: I install one from a marketplace,
-edit it, copy it to a new project, forget where the original lived. This repo
-is the one place that's allowed to be authoritative. If a skill isn't here, it
-isn't real on this laptop.
+> 127 skills across 21 categories. The source of truth for all local skills used by [cue](https://github.com/opencue/cue) profiles.
 
-Everything else (`~/.claude/skills/<name>`, `~/.codex/skills/<name>`) is a
-symlink. Edit a file under `skills/` and every agent picks it up on next load —
-no copy, no resync, no drift.
-
----
+## What's here
 
-## Layout
-
-```text
+```
 skills/
-  <category>/
-    <skill-name>/
-      SKILL.md          # the agent-facing spec
-      scripts/          # optional bundled scripts
-      references/       # optional supporting docs
-      assets/           # optional images, samples
-scripts/
-  install-local.sh      # installs to both Codex and Claude
-  install-codex.sh
-  install-claude.sh
-docs/
-  installed-sources.tsv # provenance log from last install
-  install.md
-.claude/
-  settings.json         # repo-scoped Claude Code config
+├── skills/                 The skill library
+│   ├── browser/            Playwright, screenshots
+│   ├── caveman/            Terse mode, commits, compression
+│   ├── colony/             Multi-agent coordination
+│   ├── deployment/         Coolify, Supabase, pnpm
+│   ├── design/             UI/UX, branding, SVG, Remotion
+│   ├── github/             GitHub CLI, CI fixes, auth
+│   ├── higgsfield/         AI image/video generation
+│   ├── hostinger/          DNS, domains, VPS
+│   ├── medusa/             Medusa v2 ecommerce
+│   ├── meta/               Profile management, memory, helpers
+│   ├── nvidia/             cuOpt, GPU optimization
+│   ├── obsidian/           Vault, markdown, canvas
+│   ├── orchestration/      Fleet, pipelines, workers
+│   ├── research/           Search, papers, keywords
+│   ├── review/             Code review, security, testing
+│   ├── secrets/            Envoult, credential management
+│   ├── stripe/             Payments, webhooks
+│   └── ...
+├── scripts/                Install, sync, lint scripts
+├── plugins/                Claude Code plugin definitions
+└── catalog/                Auto-generated skill index
 ```
 
----
+## Skill format
 
-## Categories
-
-| Folder            | Count | What lives here                                                  |
-| ----------------- | ----: | ---------------------------------------------------------------- |
-| `ai/`             |     3 | Anthropic SDK integration, prompt caching, Claude migration      |
-| `automation/`     |     2 | Task scheduling, cron jobs, recurring automation                 |
-| `caveman/`        |     5 | Token compression — caveman, caveman-commit, caveman-review, …   |
-| `colony/`         |     2 | Colony coordination, prompts, and handoff surfaces               |
-| `content/`        |     8 | Technical writing, copywriting, theming, docs, PDF, browser work |
-| `deployment/`     |     3 | Coolify, pnpm, Supabase                                          |
-| `design/`         |    19 | UI, UX, visual design, image-direction, brandkit, mockups        |
-| `github/`         |     6 | github CLI, gh-fix-ci, gitguardex, worktrees, branch finish      |
-| `growth/`         |    12 | CRO, analytics, retention, referrals, free-tool growth           |
-| `higgsfield/`     |     4 | Higgsfield AI — generate, soul-id, photoshoot, marketplace       |
-| `hostinger/`      |     4 | Hostinger domains, DNS, hosting, VPS                             |
-| `marketing/`      |    18 | SEO, ads, email, launch, PMM, pricing, community, RevOps-adjacent |
-| `medusa/`         |    14 | Medusa commerce, storefronts, db migrations, woocommerce import  |
-| `meta/`           |    25 | Agent meta, config, workflow, plans, doctor, hud, ask-*, trace   |
-| `obsidian/`       |     4 | Obsidian vault tooling and JSON canvas                           |
-| `orchestration/`  |    13 | autopilot, ralph, team, subagents, ultraqa, ultrawork, pipeline  |
-| `private/`        |     1 | Host- / account-specific skills (no secrets in repo)             |
-| `research/`       |    11 | autoresearch, interviews, customer/competitor research, keywords |
-| `review/`         |    11 | code/security/architecture review, debugging, TDD, verification  |
-| `stripe/`         |     2 | Stripe integration and webhooks                                  |
-
-> Skill leaf names are unique repo-wide so symlink installs never collide.
+Each skill is a directory with a `SKILL.md` file:
 
+```markdown
 ---
-
-## Install
-
-```sh
-./scripts/install-local.sh
-```
-
-The installer walks every category for `SKILL.md` files and links each parent
-folder into both:
-
-- `~/.codex/skills/<skill-name>`
-- `~/.claude/skills/<skill-name>`
-
-Existing skill folders at the destination are moved to timestamped backups
-before linking — nothing is overwritten silently.
-
-`docs/installed-sources.tsv` records where each imported skill came from at
-last import and whether duplicates were skipped.
-
+description: "When user says X, do Y"
+allowed-tools: ["tool_name"]
 ---
 
-## Adding a new skill
+# Skill Name
 
-1. Drop it under the right category as `skills/<category>/<skill-name>/`.
-2. Make sure it has `SKILL.md` with valid frontmatter (`name`, `description`).
-3. Re-run `./scripts/install-local.sh`. The installer is idempotent.
+Instructions for the model...
+```
 
-Single-skill category? Still create the folder — flat skills aren't allowed
-anymore. Better to have `deployment/coolify/` with one entry than a special
-case.
+The `description` frontmatter is what the LLM matches against to decide when to use the skill.
 
----
+## How cue uses this
 
-## Secrets
+Profiles reference skills by `category/slug`:
 
-Do not store secrets in this repo.
+```yaml
+# profiles/backend/profile.yaml
+skills:
+  local:
+    - review/code-review
+    - deployment/coolify
+    - github/gh-fix-ci
+```
 
-The `private/` category exists for skills that touch host-specific or
-account-specific surfaces. Secrets and connection details (IPs, usernames,
-tokens) **must** come from environment variables or `~/.ssh/config` aliases —
-never from committed files. The bundled scripts under `private/myvps/` enforce
-this: they fail fast if `SUPA_SCHEMA_SSH_TARGET` isn't set.
+At launch, cue symlinks these into the runtime's `skills/` directory.
 
-For WooCommerce imports, credentials live in:
+## Adding a skill
 
-```text
-~/.config/woocommerce-medusa-import/env       # mode 600
+```bash
+cue skills-new my-skill                       # scaffold
+# edit skills/<category>/my-skill/SKILL.md
+cue skills-lint my-skill                      # validate
+cue skills-test my-skill                      # test
 ```
 
----
+Or manually: create `skills/<category>/<slug>/SKILL.md` with the frontmatter format above.
 
-```
-   ─── this is mine. fork it, gut it, build your own. ───
-```
+## Categories
+
+| Category | Skills | Domain |
+|----------|--------|--------|
+| `browser` | 1 | Playwright, screenshots |
+| `caveman` | 5 | Terse mode, commits |
+| `design` | 16 | UI/UX, branding, SVG |
+| `medusa` | 14 | Ecommerce platform |
+| `meta` | 18 | Profile helpers, memory |
+| `nvidia` | 12 | GPU optimization |
+| `review` | 5 | Code quality, security |
+| `research` | 8 | Search, papers |
+
+## Related
+
+- [cue](https://github.com/opencue/cue) — the profile manager
+- [resources/mcps](../mcps/) — MCP server registry

package/resources/skills/skills/design/headless-gif-demo/SKILL.md

@@ -1,10 +1,14 @@
 ---
-description: "When user asks to record a high-quality CLI demo GIF that needs Kitty graphics protocol (real PNG icons inline) — use this headless Xvfb + Kitty + tmux + ffmpeg pipeline instead of vhs/asciinema which don't speak the Kitty protocol"
-requires_mcps: []
-allowed-tools: Bash(Xvfb:*), Bash(kitty:*), Bash(tmux:*), Bash(xdotool:*), Bash(ffmpeg:*), Bash(/usr/bin/ffmpeg:*), Read(*), Write(*)
+description: "When user asks to record a high-quality CLI demo GIF that needs Kitty graphics protocol (real PNG icons inline) — use this headless cage/Wayland + Kitty + tmux + ffmpeg pipeline instead of vhs/asciinema which don't speak the Kitty protocol. Includes auto-redaction of moving text via tesseract OCR."
+requires_mcps: [cue-tty-watch]
+allowed-tools: Bash(cage:*), Bash(weston:*), Bash(Xvfb:*), Bash(kitty:*), Bash(tmux:*), Bash(xdotool:*), Bash(wf-recorder:*), Bash(grim:*), Bash(ffmpeg:*), Bash(/usr/bin/ffmpeg:*), Bash(tesseract:*), Bash(convert:*), Read(*), Write(*), mcp__cue-tty-watch__*
 ---
 
-# Headless GIF demos with Kitty + tmux + Xvfb + ffmpeg
+# Headless GIF demos with Cage (Wayland) + Kitty + tmux + auto-redaction
+
+> Updated pipeline: prefer **cage** (headless wlroots compositor) over Xvfb when on a Wayland host.
+> Kitty graphics protocol works through wf-recorder → real brand-icon PNGs render in the GIF.
+> Tesseract OCR + `redact_video` MCP tool handles moving-text redaction automatically.
 
 When you need a demo GIF of a CLI tool that uses **Kitty graphics protocol** (e.g. cue's brand-logo PNGs in `cue optimizer`), `vhs` and `asciinema` won't work — they render in `ttyd` which doesn't speak the protocol. Logos show as garbled placeholder boxes or fall back to emoji.
 
@@ -20,15 +24,19 @@ This skill captures the working pipeline: spin up a virtual X display, run real
 
 ## Required tools
 
-| Tool | Install (apt) | Install (nix) |
+| Tool | Purpose | Install |
 |---|---|---|
-| `Xvfb` | `sudo apt install xvfb` | `nix profile install nixpkgs#xorg.xorgserver` |
-| `xdotool` | `sudo apt install xdotool` | `nix profile install nixpkgs#xdotool` |
-| `ffmpeg` with `x11grab` | `sudo apt install ffmpeg` (**not** nix-pure ffmpeg — that one lacks x11grab) | `nix profile install nixpkgs#ffmpeg-full` |
-| `kitty` | already on your box | — |
-| `tmux` | already on your box | — |
-
-**Gotcha:** verify ffmpeg has x11grab with `ffmpeg -devices | grep x11`. nix's stock `ffmpeg` doesn't ship it — use `/usr/bin/ffmpeg` (apt) or `ffmpeg-full` (nix).
+| `cage` | Headless wlroots compositor (preferred over weston — speaks `wlr-screencopy`) | `sudo apt install cage` |
+| `wf-recorder` | Wayland screen capture (needs wlroots compositor) | `sudo apt install wf-recorder` |
+| `grim` | Wayland screenshot (for preflight checks) | `sudo apt install grim` |
+| `kitty` | Terminal with graphics protocol — renders brand PNGs inline | already there |
+| `tmux` | Drives the demo via `send-keys` (works on any display protocol) | already there |
+| `/usr/bin/ffmpeg` (apt) | Used to convert mp4→gif. Nix's `ffmpeg` lacks `x11grab`/`palettegen` features sometimes — apt build is safer. | `sudo apt install ffmpeg` |
+| `tesseract` | OCR for auto-redaction. Returns bounding boxes per word/line. | `sudo apt install tesseract-ocr tesseract-ocr-eng` |
+| `ollama` + `moondream` | Optional: fast vision Q&A ("is the splash visible yet?") | `ollama pull moondream` |
+| `opencv-python`, `scenedetect`, `numpy`, `pillow` | Vision scripting in `~/.venvs/video` (or similar) | `pip install opencv-python-headless scenedetect numpy pillow` |
+
+**Fallback (X11-only systems):** `Xvfb` + `ffmpeg x11grab` + `xdotool`. Same shape, X-only verbs.
 
 ## The pipeline
 
@@ -166,3 +174,40 @@ Pathological capture (means kitty didn't render to Xvfb):
 3. Test-run; check the preflight screenshot is non-empty
 4. Iterate on timing — pickers and Claude Code splash both need 2–3 s headroom
 5. Commit both the script and the resulting GIF — re-running gives byte-identical output for a fixed tape
+
+## Auto-redaction with the cue-tty-watch MCP
+
+When the demo captures sensitive moving text (email in a splash card that scrolls as content is added below), don't fight per-frame `drawbox` coordinates by hand. The `cue-tty-watch` MCP (autoloaded on every cue profile via `core` inheritance) wraps tesseract + scenedetect + moondream so the agent can do this automatically:
+
+| MCP tool | Backed by | Use when |
+|---|---|---|
+| `screenshot(display)` | `xwd` / `grim` | "What's on the headless display right now?" |
+| `tmux_pane(socket, session)` | `tmux capture-pane -p` | "What text did the inner shell render?" |
+| `send_keys_tmux(...)` | `tmux send-keys` | Drive the demo non-interactively |
+| `find_text(image, query)` | `tesseract` | Get `{x,y,w,h}` for every occurrence of a substring in a frame |
+| `ask_about_image(image, q)` | `ollama` + `moondream` | Fast yes/no-style vision Q&A ("did the splash render?") |
+| `detect_scenes(video)` | `scenedetect` (`ContentDetector`) | Find scene-cut timestamps automatically |
+| **`redact_video(in, out, query)`** | All of the above | **One call**: walks the video, OCRs each sampled frame, finds `query`, builds per-frame drawbox+drawtext filter, encodes the output |
+
+`redact_video` is the killer. The "email scrolls up as Claude responds and my fixed drawbox is in the wrong place" problem reduces to:
+
+```jsonc
+// MCP call
+{
+  "tool": "redact_video",
+  "input_path": "/tmp/cue-demo-raw.mp4",
+  "output_path": "/tmp/cue-demo-redacted.mp4",
+  "query": "webubusiness",
+  "label": "[ account info · redacted ]",
+  "samples_per_second": 4,
+  "pad": 6
+}
+```
+
+Then convert the redacted mp4 → gif with the standard 2-pass palette pipeline.
+
+Why this beats the manual approach:
+- No more 3-phase time-gated drawboxes that leak at transitions
+- Adapts automatically to any future demo (different fonts, scroll speeds, splash layouts)
+- Handles boundaries cleanly — text disappears one frame, box disappears the next
+- Works for ANY moving sensitive text, not just the email row

package/resources/skills/skills/event-design/wedding-invitations/SKILL.md

@@ -0,0 +1,43 @@
+---
+name: wedding-invitations
+description: Use when designing a wedding invitation, save-the-date, or RSVP card from a conversation — bespoke HTML rendered to a print-ready PNG, any language, any aesthetic, fully local. Pointer to upstream wyx-sg/wedding-invitation-skill.
+allowed-tools: Bash(chromium:*), Bash(google-chrome:*), Bash(microsoft-edge:*), Bash(node:*), Bash(git:*)
+---
+
+# Wedding Invitations — bespoke, local, multilingual
+
+[`wyx-sg/wedding-invitation-skill`](https://github.com/wyx-sg/wedding-invitation-skill) — designs a one-off wedding invitation from a conversation. Outputs a print-ready PNG (1080×1440 portrait or 1080×1920 9:16 poster). Renders locally via a headless Chromium browser; no uploads, no cloud, no telemetry.
+
+## When to use
+- "Help me make a wedding invitation"
+- "Design a save-the-date in [language]"
+- Save-the-dates, RSVP cards, programs, menus, place cards — anything the same HTML→PNG pipeline can produce
+- Multilingual collateral (English, Chinese, Spanish, Japanese, Korean, or any combination)
+- Aesthetic directions covered by the upstream gallery: new-chinese, wabi-sabi, art-deco, morandi, modern-minimal, mediterranean, retro-poster, etc.
+
+## How
+1. **Talk** — the skill asks for language(s), names, date, venue, style preference
+2. **Preview** — aesthetic directions shown visually in your browser
+3. **Design** — Claude writes a unique HTML template from scratch
+4. **Iterate** — natural-language tweaks ("bigger font" / "softer color" / "swap the photo")
+5. **Export** — one command screenshots the HTML → high-res PNG
+
+## Install
+```bash
+git clone https://github.com/wyx-sg/wedding-invitation-skill \
+  ~/.claude/skills/wedding-invitation
+```
+Then invoke via `/wedding-invitation` in Claude Code, or just ask: "help me make a wedding invitation."
+
+## Prerequisites
+- `node` 18+
+- A Chromium-family browser: `chromium`, `google-chrome`, or `microsoft-edge` (any one suffices)
+- `git` for the install
+
+The upstream `render.js` auto-locates whichever browser you have. On Linux: `cue cli install chromium`. On macOS: `brew install --cask chromium` (or any Chrome already installed works). On Windows: Edge ships with the OS.
+
+## Notes
+- **Privacy**: photos, names, addresses stay on your machine. The only network requests are Google Fonts CDN loads from your browser during HTML preview — font URLs only, nothing about you.
+- **Not a template gallery**: each invitation is designed fresh from your prompt. The 20 gallery examples are showcases, not picks.
+- **Other agents**: the skill is Claude-Code-first but works with Codex CLI, Cursor, Aider, Gemini CLI — tell the agent "read SKILL.md and help me make a wedding invitation."
+- **Sibling collateral**: this skill specializes in invitations but the HTML→PNG pipeline generalizes — for programs/menus/place cards, the same workflow applies; ask Claude to adapt the template.

package/resources/skills/skills/meta/acpx/SKILL.md

@@ -0,0 +1,78 @@
+---
+name: acpx
+description: Delegate work to another coding agent (codex, claude, pi, openclaw, gemini, cursor, copilot, droid, etc.) over the Agent Client Protocol via acpx. Use when the user says "delegate to <agent>", "run codex", "use claude code", "have <agent> do X", "spawn a sub-agent", "agent-to-agent", mentions ACP, or when the work belongs in a different harness than the one you're in. Prefer this over PTY scraping or `tmux send-keys` to a foreign agent.
+---
+
+# acpx — talk to other coding agents over ACP
+
+`acpx` is a headless CLI that lets one agent drive another over the **Agent
+Client Protocol (ACP)**. Persistent sessions, prompt queueing, structured
+tool-call output, soft-close lifecycle. You get a real conversation with the
+target agent instead of scraping its terminal.
+
+⚠️ acpx is in alpha; the surface may shift. Pin via `npm install -g acpx@latest`
+or `npx acpx@latest` for one-offs. State lives in `~/.acpx/`.
+
+## When to reach for it
+
+- User asks you to **delegate** a task to another agent ("get codex to fix X",
+  "have claude refactor Y").
+- You're in one harness and the task is clearly a better fit for another
+  (e.g. you're in Claude Code and the user wants the Codex CLI to run the
+  task because of model availability or workflow preference).
+- The user wants **parallel work in the same repo** by different agents
+  without stomping on each other.
+- You'd otherwise be tempted to `tmux send-keys`, `expect`, or screen-scrape
+  another agent's TUI — stop, use acpx instead.
+
+Do **not** use acpx for the agent you're already running inside. Use your
+native tools.
+
+## Setup (one-time)
+
+```
+npm install -g acpx@latest        # global (faster)
+# or
+npx acpx@latest <agent> "<prompt>"  # no install
+```
+
+Underlying coding-agent CLIs (`codex`, `claude`, `gemini`, etc.) must already
+be installed and authenticated separately — acpx is a client, not a bundler.
+Run `cue cli install acpx` to install via the cli recipe registry.
+
+## Core commands
+
+```
+acpx codex sessions new                   # create a session for this dir
+acpx codex 'fix the failing tests'        # run a prompt
+acpx codex -s api 'paginate the endpoint' # parallel named session
+acpx codex sessions list                  # see what's open
+acpx codex status                         # is it running / idle / dead
+acpx codex cancel                         # cooperative cancel (sends ACP cancel)
+acpx codex sessions close                 # soft-close (keeps history)
+acpx codex exec 'one-shot summary'        # stateless, no saved session
+```
+
+Substitute `codex` with `claude`, `pi`, `openclaw`, `gemini`, `cursor`,
+`copilot`, `droid`, `iflow`, `kilocode`, `kimi`, `kiro`, `opencode`, `qoder`,
+`qwen`, `trae`. For custom ACP servers: `acpx --agent './bin/my-acp' '...'`.
+
+## Useful flags
+
+- `--no-wait` — queue prompt, return immediately (fire-and-forget).
+- `--format json` — NDJSON event stream, good for automation/jq pipelines.
+- `--format quiet` — final assistant text only.
+- `--approve-all` / `--approve-reads` / `--deny-all` — permission gates.
+- `--cwd <path>` — run against a different working directory.
+- `--timeout <s>` — wall-clock cap on a single prompt.
+- `--ttl <s>` — keep queue owner alive between prompts (default 300).
+- `--file <path>` / stdin — load prompt body from file or pipe.
+
+## Full reference
+
+- Skill source: <https://raw.githubusercontent.com/openclaw/acpx/main/skills/acpx/SKILL.md>
+- CLI reference: <https://raw.githubusercontent.com/openclaw/acpx/main/docs/CLI.md>
+- Built-in agent list: <https://github.com/openclaw/acpx/blob/main/agents/README.md>
+
+When something's not obvious, read those — they're the canonical source and
+they evolve faster than this stub.