Poco is a self-hosted collaborative AI workspace designed as an alternative to OpenClaw, with chat channels, task management, file handling, and a web interface. Its Claude Code–based agent works inside an isolated sandbox and can use project files, skills, MCP tools, and sub-agents.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add poco-ai/poco-claw --skill minimax-multimodal-toolkitgit clone --depth 1 https://github.com/poco-ai/poco-clawWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/poco-ai/poco-claw/minimax-multimodal-toolkit)<a href="https://agentmods.dev/skills/poco-ai/poco-claw/minimax-multimodal-toolkit"><img src="https://agentmods.dev/badge/skills/poco-ai/poco-claw/minimax-multimodal-toolkit/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/poco-ai/poco-claw/minimax-multimodal-toolkit"><img src="https://agentmods.dev/badge/skills/poco-ai/poco-claw/minimax-multimodal-toolkit.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 3 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Tool Misuse · line 29 Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).Fix: Validate all tool parameters against an allowlist. Reject dangerous parameter values (shell=True, --force, -rf /) and use safe defaults.
- medium Excessive Agency · line 247 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
- medium Excessive Agency · line 250 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00155 | $0.07968 |
| Opus 5.5 | $0.00062 | $0.03187 |
| Sonnet 5.5 | $0.00031 | $0.01594 |
| Haiku 4.5 | $0.00015 | $0.00797 |
Grade C, and why
minimax-multimodal-toolkit scanned grade C with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 1mo ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
4. Intermediate/temp files (segment audio, video segments, extracted frames) are automatically placed in `minimax-output/tmp/`. They can be cleaned up when no longer needed: `rm -rf minimax-output/tmp` Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
No Python or pip required — all scripts are pure bash using `curl`, `ffmpeg`, `jq`, and `xxd`. Copies of this mod
1 near-identical copy found in the catalogue:
- minimax-multimodal-toolkit — 100% identical, 0 lines differ
How it starts
The opening of the file, as written. The whole thing — 675 lines — stays where its author put it; the contents beside it link to each section on GitHub.
MiniMax Multi-Modal Toolkit
Generate voice, music, video, and image content via MiniMax APIs — the unified entry for MiniMax multimodal use cases (audio + music + video + image). Includes voice cloning & voice design for custom voices, image generation with character reference, and FFmpeg-based media tools for audio/video format conversion, concatenation, trimming, and extraction.
Output Directory
All generated files MUST be saved to minimax-output/ under the AGENT'S current working directory (NOT the skill directory). Every script call MUST include an explicit --output / -o argument pointing to this location. Never omit the output argument or rely on script defaults.
Rules:
- Before running any script, ensure
minimax-output/exists in the agent's working directory (create if needed:mkdir -p minimax-output) - Always use absolute or relative paths from the agent's working directory:
--output minimax-output/video.mp4 - Never
cdinto the skill directory to run scripts — run from the agent's working directory using the full script path - Intermediate/temp files (segment audio, video segments, extracted frames) are automatically placed in
minimax-output/tmp/. They can be cleaned up when no longer needed:rm -rf minimax-output/tmp
Prerequisites
brew install ffmpeg jq # macOS (or apt install ffmpeg jq on Linux)
bash scripts/check_environment.sh
No Python or pip required — all scripts are pure bash using curl, ffmpeg, jq, and xxd.
API Host Configuration
MiniMax provides two service endpoints for different regions. Set MINIMAX_API_HOST before running any script:
| Region | Platform URL | API Host Value |
|---|---|---|
| China Mainland（中国大陆） | https://platform.minimaxi.com | https://api.minimaxi.com |
| Global（全球） | https://platform.minimax.io | https://api.minimax.io |
# China Mainland
export MINIMAX_API_HOST="https://api.minimaxi.com"
# or Global
export MINIMAX_API_HOST="https://api.minimax.io"
What ships with it
15 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- references/image-api.md 3.6 KB
- references/music-api.md 1.7 KB
- references/tts-guide.md 3.1 KB
- references/tts-voice-catalog.md 32 KB
- references/video-api.md 4.8 KB
- references/video-prompt-guide.md 3.6 KB
- scripts/check_environment.sh 3.5 KB runs code
- scripts/image/generate_image.sh 12 KB runs code
- scripts/media_tools.sh 18 KB runs code
- scripts/music/generate_music.sh 9.3 KB runs code
- scripts/tts/generate_voice.sh 27 KB runs code
- scripts/video/add_bgm.sh 7.7 KB runs code
- scripts/video/generate_long_video.sh 17 KB runs code
- scripts/video/generate_template_video.sh 6.8 KB runs code
- scripts/video/generate_video.sh 11 KB runs code
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 1mo ago First seen · 675 lines · 155 tokens per session scan C d3db8a843cb0
minimax-multimodal-toolkit is a skill published in the GitHub repository poco-ai/poco-claw (1,356 stars, last pushed 7d ago), licensed MIT. It adds 155 tokens to every session and 7,968 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it C with 2 findings (recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
lobu-operator
Contribute safely to the Lobu monorepo: worktrees, package rules, red-to-green fixes, validation gates, SDK-first operations, PRs, and rollout checks.
company-agent-infrastructure
Use when designing or building internal/company AI agents that need shared context across company systems, durable organizational memory across sessions or agents, user-scoped permissions, approvals, audit, or governed actions. Also use when deciding whether ordinary MCP tools, RAG, or local agent memory are enough…
lobu-builder
Use when working inside a Lobu project generated by @lobu/cli or any repository centered on lobu.config.ts, AGENTS.md, agent prompt files, local skills, and evals. This skill helps a coding agent inspect the right files, make Lobu-native changes, keep the stack runnable, and validate semantics with chat tests and…
crm-ops
How to operate the Lobu funnel CRM — create and enrich leads, log interactions, advance funnel stages, open and update pilots, and produce the weekly digest. Use whenever the task touches the pipeline.
deliveroo-order
Turn collected lunch orders into a clean per-person order list for a human to place on Deliveroo. Use in step 2 of the lunch run, after orders are collected. The live menu is fetched automatically by the lobu-team-lunch-finalize reaction (via the Owletto Chrome extension) — this skill never places an order or touches…
lobu
Use Lobu MCP for shared, permission-aware company context, durable organizational memory, and governed actions. Trigger when the user asks what the organization knows, needs context from connected company systems, wants to preserve a durable fact or decision, or needs to discover and use Lobu SDK capabilities.