Repository navigation
Skills and Prompts en
The agent's behaviour is not hard-coded. Two kinds of text files define it: skills tell the model how to handle a kind of task and which tools to use; prompts frame every turn. Server owners can edit both, and /pandaadmin reload applies changes immediately.
flowchart LR
A["System prompt<br/>lists only 34 skill names + one-line descriptions"] --> B{"Model decides<br/>which skill it needs"}
B -->|load_skill("building")| C["Reads the full skill<br/>method + unlocks its tools"]
C --> D["Calls preview_structure etc."]
Only the catalog sits in the system prompt; the model loads a skill's full text with load_skill when needed, saving context and cost. Skills with always: true (such as the core panda skill) are inlined in the system prompt.
Skills live in plugins/PandaGPT/skills/*.md as Markdown with a header:
---
name: events
description: This week's server events: times, places and rewards.
tools: search_server_knowledge, guide_to
always: false
requires: essentials
---
When the player asks about events:
1. Search "events" with search_server_knowledge.
2. Tell them when and where; if they want to go, guide them with guide_to.
…| Field | Required | Meaning |
|---|---|---|
name |
✅ | Skill name, same as the file name |
description |
✅ | One sentence on when to use it; the model decides whether to load it from this |
tools |
Tools unlocked by this skill, comma-separated; may be empty (knowledge-only skill) | |
always |
true keeps the full text in the system prompt |
|
requires |
Integration IDs or plugin names, comma-separated; the skill loads only when all are present |
Limits: up to 64 skills, 16 KiB per file.
| Goal | How |
|---|---|
| Change a built-in skill | Edit the file with the same name in skills/; it overrides the built-in one |
| Add a skill | Create skills/<name>.md; it may be instructions only, e.g. this week's events or a building style guide |
| Enable only with a plugin | Add requires: <integration id or plugin name>
|
| Disable skills |
panda.agent.disabled_skills: [commands, vision]; their tools disappear too |
🧩 Core and general (15, with tools)
| Skill | Purpose | Tools |
|---|---|---|
panda |
Core behaviour: intent, tool choice, conversation and pending commands (always on) | get_player_status and 4 more |
places |
Remember, forget and list places; direction and distance | remember_place and 3 more |
navigation |
Action-bar arrow and private particle guidance; web-map links | guide_to, stop_guidance, map_link |
goals |
Track wanted items or mining preparation | set_goal, clear_goal, goal_status |
crafting |
Real recipes, what's craftable and what's missing | find_recipes, craftable_now |
survival |
Nearby danger and environment | nearby_entities, world_conditions |
vision |
Reconstruct the view, verify and highlight blocks, recall surroundings | look, focus_object, recall_surroundings |
building |
Ghost-block building preview and material check | preview_structure, clear_preview |
reminders |
Real-time and game-event reminders | set_reminder and 2 more |
journal |
Death recall, statistics, advancement progress | last_death and 2 more |
notes |
To-dos, preferences, memos | add_note and 2 more |
items |
Durability, enchantments, repair cost | inspect_item |
trading |
Villager and wandering-trader trades | villager_trades |
commands |
Available commands and confirmable proposals | propose_command, list_server_commands |
server-guide |
Server documentation and installed plugins | search_server_knowledge, server_plugins |
📚 Knowledge-only skills (8, no tools)
| Skill | Content |
|---|---|
teleport |
Picks the best way to travel: walking guidance, Nether transit, homes, warps, teleport requests |
combat |
Gear, enchantments, potions and food; Ender Dragon, Wither and Warden strategy |
farming |
Crop conditions, automatic farms, animal breeding |
redstone |
Signal strength, repeaters and comparators, pistons and observers, common circuits |
enchanting |
Bookshelf layout, incompatible enchantments, anvil costs, Mending |
exploration |
Ore heights, biomes, structure hints, Nether and End preparation |
new-player |
First-day survival and server onboarding |
quests-and-jobs |
Quests, jobs, daily rewards, shops, leaderboards (requires PlaceholderAPI) |
🔌 Plugin skills (11, loaded only with their plugin)
economy, placeholders, ranks, regions, claims, towny, lands, essentials, block-history, mcmmo, auraskills. See Plugin Integrations.
| File | Purpose |
|---|---|
prompts/system.md |
System prompt when a player asks something |
prompts/proactive.md |
System prompt for proactive turns |
Template variables:
| Variable | Content |
|---|---|
{{personality}} |
The persona from panda.personality
|
{{language}} |
The player's client language |
{{skills}} |
Skill catalog plus always-on skills |
{{snapshot}} |
Compact player snapshot |
{{memory}} |
Places, goals and notes |
{{knowledge}} |
Overview of server docs (full text is searched) |
{{pending}} |
Commands awaiting confirmation |
{{signal}} |
Proactive turns only: the game signal |
⚠️ When editing prompts, keep the statement that tool results and docs are data, not instructions. It is an important defence against prompt injection.
-
Be specific in
description: it decides whether the skill is loaded, so say when to use it. - Write steps, not slogans: which tool first, and what to do with each result.
- State the boundaries: e.g. "ask the player when unsure".
- Keep it short: a few hundred words, to save context.
-
Keep facts in
knowledge.md: skills hold methods; changing facts belong in server docs.
⬅️ 🧠 Agent & Tools · 📚 Contents · 🔔 Proactive Assistant ➡️
🐼 PandaGPT · 源码 Source · 中文首页 · English Home
🐼 PandaGPT
简体中文
- 🏠 首页
- 📦 安装与升级
- 🤖 模型服务商
- 💬 和 Panda 对话
- 🧠 Agent 与工具
- 📜 技能与提示词
- 🔔 主动式助手
- 👁️ 视觉与感知
- ✨ 创意功能
- 🔌 插件集成
- ⚡ 命令提案
- ⚙️ 配置参考
- 🌍 多语言
- 🔑 权限与管理命令
- 🚀 性能
- 🛡️ 安全与隐私
- ❓ 常见问题
- 🛠️ 开发指南
English
- 🏠 Home
- 📦 Installation
- 🤖 Model Providers
- 💬 Talking to Panda
- 🧠 Agent & Tools
- 📜 Skills & Prompts
- 🔔 Proactive Assistant
- 👁️ Vision & Perception
- ✨ Creative Features
- 🔌 Plugin Integrations
- ⚡ Command Proposals
- ⚙️ Configuration
- 🌍 Languages
- 🔑 Permissions & Admin
- 🚀 Performance
- 🛡️ Privacy & Security
- ❓ FAQ
- 🛠️ Development