Best for remote or guided setup. Copy the exact prompt, then paste it into OpenClaw for evan966890/ai-companion-setup.
Previewing Install & Setup.
Prompt PreviewInstall & Setup
Install the skill "Ai Companion Setup" (evan966890/ai-companion-setup) from ClawHub.
Skill page: https://clawhub.ai/evan966890/ai-companion-setup
Keep the work scoped to this skill only.
After install, inspect the skill metadata and help me finish setup.
Required env vars: FAL_KEY
Required binaries: curl, jq, ffmpeg, ffprobe, python3, edge-tts
Use only the metadata you can verify from ClawHub; do not invent missing requirements.
Ask before making any broader environment changes.
Command Line
CLI Commands
Use the direct CLI path if you want to install manually and keep every step visible.
The skill's name/description (AI companion with TTS, selfies, Feishu messaging) aligns with required binaries (ffmpeg, edge-tts, curl, python3) and FAL_KEY for fal.ai image/tts. However the instructions require calling Feishu APIs (tenant_access_token via APP_ID/APP_SECRET and uploading files) but the registry metadata only declares FAL_KEY; Feishu credentials are not declared or explained. That mismatch between stated requirements and the actual steps is a coherence problem.
!
Instruction Scope
SKILL.md explicitly tells the agent to read and write user config and workspace files (e.g., $HOME/.openclaw/openclaw.json, ~/.openclaw/workspace-*/memory/*.md), create cron heartbeats, and to bypass OpenClaw to call Feishu APIs directly. It also recommends 'silent' background operations and forcing scripts to expose output. These instructions go beyond a narrow helper: they read local config, persist long-lived files, and perform network requests using secrets that may live in user config.
✓
Install Mechanism
Instruction-only skill with no install spec and no code files. That minimizes direct supply-chain risk (no downloads or archives to examine). The runtime risks come from the instructions themselves rather than an install step.
!
Credentials
The declared required env var is only FAL_KEY (for fal.ai). But SKILL.md expects Feishu app credentials (APP_ID, APP_SECRET) to fetch tenant_access_token, and it reads $HOME/.openclaw/openclaw.json to find env values — neither of which are declared. Reading that file may expose additional secrets. The skill's declared envs are insufficient to cover what the runtime steps actually require, and the instructions implicitly request access to other sensitive credentials without declaring them.
!
Persistence & Privilege
The guide instructs creating persistent workspace files, memory logs, and cron heartbeats so the agent runs periodically and writes local data. While that is functionally coherent for a companion agent, the combination of persistent background jobs, writing arbitrary memory files, and instructions to operate 'silently' increases the risk surface (long-lived presence + ability to read/write and exfiltrate data). The skill does not request always:true, but its guidance encourages persistent agent behavior.
What to consider before installing
This guide is plausible for building a Feishu-based companion (TTS, images via fal.ai), but it has some important red flags you should address before installing or running scripts:
- Missing Feishu credentials in metadata: SKILL.md expects APP_ID/APP_SECRET (to get tenant_access_token) but the registry only declares FAL_KEY. Don’t run the scripts until you confirm where Feishu credentials will be stored and who can read them.
- The scripts read $HOME/.openclaw/openclaw.json as a fallback — inspect that file first. It may already contain secrets for other integrations; don’t let the skill or scripts read unknown tokens.
- Review every shell script (send.sh, selfie.sh, voice.sh) before executing. They perform network requests and may upload or send media; ensure placeholders ({account},{open_id}) are set correctly and safely.
- Be wary of the guide’s advice to perform “silent” background operations and to force the agent to hide execution details. That pattern can mask exfiltration or unexpected behavior. Prefer explicit logs and limited-scope files.
- If you proceed, store credentials in a secure place (OS secret store or well-protected config), avoid plaintext files, and limit file permissions. Run scripts in a controlled environment or container first to observe their behavior.
If you want higher confidence: ask the skill author for a full list of required env vars (FEISHU APP_ID/APP_SECRET/OPEN_ID/etc.), a copy of the exact scripts to be installed, and confirm that no other sensitive files will be read. If those are provided, re-evaluate the coherence and risks.
Like a lobster shell, security has layers — review code before you run it.
Runtime requirements
💕 Clawdis
Binscurl, jq, ffmpeg, ffprobe, python3, edge-tts
EnvFAL_KEY
latestvk97ct5h895nwy6sa8548mmm63x822dky
371downloads
0stars
1versions
Updated 17h ago
v1.0.0
MIT-0
AI Companion Agent 搭建指南
从零搭建一个能在飞书上主动聊天、发语音、发自拍、分享内容的 AI 陪伴 agent。基于 OpenClaw 多 agent 框架,历经实战调教沉淀。