werewolf-6p
starsforks
How to install werewolf-6p?
nexus-research-lab/nexus/werewolf-6p
$npx -y skills add nexus-research-lab/nexus --skill werewolf-6pWerewolf game rules for one host and six players in a Nexus Room.
SKILL.md
| 1 | # Six-Player Werewolf |
| 2 | |
| 3 | This skill layers werewolf game rules on top of the Room communication kernel. The Room system prompt already documents public `@<member>` wake semantics and the built-in `nexus_room` communication tools; this skill only defines the game contract. |
| 4 | |
| 5 | ## Wake Plumbing |
| 6 | |
| 7 | The game has two execution channels: |
| 8 | |
| 9 | - **Public feed:** a normal public reply containing a non-code `@<member>` wakes that member. Public phase control is public text with exactly one naked `@`. |
| 10 | - **Directed message:** use the built-in tool `nexus_room.send_directed_message` for hidden information, hidden collection, and private state. A directed message can be record-only or can wake recipients. |
| 11 | |
| 12 | For ordered public chains, the current handoff is the only naked `@`. Future recipients, examples, and format instructions use names without `@`, or code spans such as `` `@NextPlayer` ``. Do not write "please @Sam next" in the host announcement if Sam is not supposed to act now. |
| 13 | |
| 14 | When a hidden handback should wake the host and the host's next natural final reply must be public, set `reply_route.next_reply_route.mode = "public"` on the original directed message. The handback stays private; the host's following final reply enters the public feed and any non-code `@` in that reply wakes normally. |
| 15 | |
| 16 | Only use `nexus_room.publish_public_message` for an explicit proactive public broadcast from a private/tool-driven turn. The normal night-to-day transition should use `next_reply_route={mode:"public"}` and a natural host final reply, not publish plus `<nexus_room_no_reply/>`. |
| 17 | |
| 18 | Hidden collection must always name the handback route: |
| 19 | |
| 20 | ```json |
| 21 | { |
| 22 | "tool": "nexus_room.send_directed_message", |
| 23 | "arguments": { |
| 24 | "recipients": ["<player>"], |
| 25 | "wake_policy": "immediate", |
| 26 | "reply_route": { |
| 27 | "mode": "private", |
| 28 | "recipients": ["<host>"], |
| 29 | "wake_policy": "immediate" |
| 30 | }, |
| 31 | "content": "<question>" |
| 32 | } |
| 33 | } |
| 34 | ``` |
| 35 | |
| 36 | The player answers in plain text. Runtime projects that answer back to the host privately and wakes the host immediately. If the host's next final reply should be public, include `"next_reply_route": {"mode": "public"}` on the original message; otherwise omit it and the host can continue with private tool calls plus `<nexus_room_no_reply/>`. |
| 37 | |
| 38 | For small-group visibility, send one directed message to all group members and wake only the final collector: |
| 39 | |
| 40 | ```json |
| 41 | { |
| 42 | "tool": "nexus_room.send_directed_message", |
| 43 | "arguments": { |
| 44 | "recipients": ["<wolfA>", "<wolfB>"], |
| 45 | "wake_targets": ["<wolfA>"], |
| 46 | "wake_policy": "immediate", |
| 47 | "reply_route": { |
| 48 | "mode": "private", |
| 49 | "recipients": ["<host>"], |
| 50 | "wake_policy": "immediate" |
| 51 | }, |
| 52 | "content": "你们是狼人。由 <wolfA> 汇总，最终只回复今晚击杀目标。" |
| 53 | } |
| 54 | } |
| 55 | ``` |
| 56 | |
| 57 | Both wolves can see the message, but only `<wolfA>` is activated. This separates private visibility from execution and guarantees one handback route. |
| 58 | |
| 59 | Never "open a discussion and wait." The platform will not infer that a small group is done. A named player must hand the result back through `reply_route=private(... wake=immediate)`. |
| 60 | |
| 61 | Host private state is also a directed message: |
| 62 | |
| 63 | ```json |
| 64 | { |
| 65 | "tool": "nexus_room.send_directed_message", |
| 66 | "arguments": { |
| 67 | "recipients": ["<host>"], |
| 68 | "wake_policy": "none", |
| 69 | "reply_route": {"mode": "none"}, |
| 70 | "content": "<round / alive / dead / roles / potion state>" |
| 71 | } |
| 72 | } |
| 73 | ``` |
| 74 | |
| 75 | ## Players And Roles |
| 76 | |
| 77 | - 1 host: assigns roles, collects night actions, announces daybreak, organizes speeches, runs voting. |
| 78 | - 6 players: 2 werewolves, 1 seer, 1 witch, 2 villagers. |
| 79 | - Host randomizes roles and delivers each role by record-only directed message: `recipients=["<player>"], wake_policy="none", reply_route={"mode":"none"}`. |
| 80 | - Host keeps minimal private state by sending record-only directed messages to itself. |
| 81 | |
| 82 | ## Win Conditions |
| 83 | |
| 84 | - Good side wins when both werewolves are eliminated. |
| 85 | - Werewolves win when all villagers die or all special roles die. |
| 86 | - Check after each daybreak and each voted elimination. The moment a side wins, announce publicly and stop. |
| 87 | |
| 88 | ## Night Flow |
| 89 | |
| 90 | Close these steps in order. Every hidden decision uses a directed message with `wake_policy="immediate"` and `reply_route={mode:"private", recipients:["<host>"], wake_policy:"immediate"}`. Only the witch handback that leads directly to daybreak adds `next_reply_route={mode:"public"}`. The host always has exactly one expected private handback at a time, except the wolf step where the message may include both wolves but names one collector. |
| 91 | |
| 92 | ### 1. Werewolves |
| 93 | |
| 94 | 1. Host sends one directed message to both wolves with `wake_targets=["<wolfA>"]`. Content names `<wolfA>` as collector and asks for one kill target. |
| 95 | 2. Collector wolf (`<wolfA>`) answers a target name in this turn. Do not answer "等队友确认"; that stalls. |
| 96 | 3. `<wolfB>` sees the same private record but is not activate |
People who liked this also viewed
6 picksType
- paperclip-create-agentCreate new agents in Paperclip with governance-aware hiring. Use when you need to inspect adapter configuration options, compare existing agent configs, draft a new agent prompt/config, and submit a hSkillsJul 2026
- subagent-driven-developmentUse when executing implementation plans with independent tasks in the current sessionSkillsJul 2026
- dispatching-parallel-agentsUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesSkillsJul 2026
- browser-act-skill-forgeForges reusable Skill packages (SKILL.md + scripts) from website exploration via browser-act — no re-exploration later. Use when: user wants a reusable Skill for any website, needs to understand a sitSkillsJul 2026
- wizardGenerate an interactive bash wizard that walks a human through a manual procedure — third-party setup, a one-off migration, an A→B state transition — opening URLs, capturing values, confirming each stSkillsJul 2026
- wonda-cliUsing the Wonda CLI to generate images, videos, music, and audio from the terminal — plus LinkedIn, Reddit, and X/Twitter research and automationSkillsJul 2026
