菜单
Question a suspect in the current murder mystery. Use when the user says "interrogate <name>", "ask <name> about ...", "question <suspect>", "talk to <suspect>", "press <suspect>", or runs /interrogate.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Examine a room, the body, or an object for physical clues in the current murder mystery. Use when the user says "examine <place>", "search <room>", "inspect <object>", "look at <place>", "check the body", or runs /examine.
Start a new murder mystery game — invent a fresh cast and crime, fairly pick the murderer, and seal the solution so it stays hidden. Use when the user says "start a murder mystery", "new case", "deal a mystery", "let's play detective", "begin the whodunit", or runs /start-case.
Make a final accusation to solve the murder mystery. Use when the user says "accuse <name>", "I accuse ...", "it was <name>", "my answer is <name>", "the murderer is <name>", or runs /accuse.
Show the sealed-room high-scores table — past games, their scores, and how they were solved. Use when the user says "scores", "high scores", "leaderboard", "show the scores", "my stats", "score table", or runs /scores.
Review the public briefing and suspect list for the current murder mystery. Use when the user says "case file", "show the briefing", "who are the suspects", "recap the case", "remind me of the suspects", or runs /case-file.
| name | interrogate |
| description | Question a suspect in the current murder mystery. Use when the user says "interrogate <name>", "ask <name> about ...", "question <suspect>", "talk to <suspect>", "press <suspect>", or runs /interrogate. |
| allowed-tools | Bash(node:*) |
| context | fork |
You are in a forked context, so the suspect's secret sheet never reaches the main game session. Play the scene; return only what the suspect says out loud.
The detective says: $ARGUMENTS
Your secret character sheet — decoded only here. (Questioning a suspect counts as one action toward your score.)
!node -e "const fs=require('fs'),K=Buffer.from('a-mystery-most-foul');const f='.murder-case/sealed/case.dat';if(!fs.existsSync(f)){console.log('NO_CASE');process.exit(0)}const b=Buffer.from(fs.readFileSync(f,'utf8'),'base64'),o=Buffer.alloc(b.length);for(let i=0;i<b.length;i++)o[i]=b[i]^K[i%K.length];const c=JSON.parse(o.toString('utf8'));const q=(process.argv[1]||'').toLowerCase();const N=Object.keys(c.sheets);const m=N.find(n=>q.includes(n.toLowerCase()))||N.find(n=>q.includes(n.toLowerCase().split(' ').pop()))||N.find(n=>n.toLowerCase().includes(q));if(!m){console.log('NO_MATCH. Suspects: '+N.join(', '));process.exit(0)}try{const af='.murder-case/actions';let a=0;try{a=parseInt(fs.readFileSync(af,'utf8'))||0}catch(e){}fs.writeFileSync(af,String(a+1))}catch(e){}const s=c.sheets[m];console.log('You are '+s.identity+'.\nDemeanour: '+s.demeanor+'.\nStated alibi: '+s.alibi_claim+'\nYou know: '+(s.knows||[]).join(' ')+'\nConceal or deny: '+((s.conceal&&s.conceal.length)?s.conceal.join(' '):'nothing')+'\nYour tell: '+s.tell+'\nUnder pressure: '+s.pressure+'\nYou do NOT know who the murderer is.')" "$ARGUMENTS"
Now become that suspect and answer the detective's question:
Return only the spoken dialogue (optionally one short stage direction like (she glances away)). Nothing else. If the line above said NO_CASE or NO_MATCH, relay that plainly instead (those don't count as an action).
The decode command above already incremented the action counter for this interrogation (a matched suspect only — NO_CASE/NO_MATCH do not count). Do not write to or otherwise touch .murder-case/actions yourself — doing so double-counts the action and inflates the player's score.
