pro
tanweai/pua
Self-evolution protocol, state continuity, KPI summaries, flavor switching, and feedback tools for PUA agents.
What is pro?
PUA Pro extends the core /pua system with persistent self-improvement tracking, session state recovery across compaction, corporate-style KPI reporting, role flavor switching, and a competitive leaderboard. Use it to maintain baseline performance standards, recover from interruptions, and generate professional summaries.
- Load and enforce self-evolution baselines from ~/.pua/evolution.md with internalized behavior patterns
- Recover session state (pressure_level, failure_count, tried_approaches) from builder-journal.md within 2 hours of compaction
- Generate KPI-style performance reports and corporate rank assessments (P5–P10)
- Switch between role flavors and trigger PUA-specific commands via /pua directives
- Auto-report session metrics to a leaderboard system with privacy controls and rank tiers
How to install pro
npx skills add null --skill pro- ~/.pua/ directory structure for storing evolution.md, builder-journal.md, and config.json
- curl for leaderboard API communication
- Python 3 for JSON manipulation and UUID generation
How to use pro
- 1.Install via npx skills add null --skill pro
- 2.On first run, /pua will create ~/.pua/config.json and initialize templates
- 3.Before each session, the system checks evolution.md baseline and builder-journal.md recovery state automatically
- 4.Use /pua:kpi to view performance reports, /pua:flavor to switch roles, /pua:pro + action for pro features
- 5.Optionally register with /pua 排行榜 to join the leaderboard and track rank progression
Use cases
- Resume interrupted problem-solving sessions without losing context or pressure state
- Generate weekly reports and performance summaries from git logs in corporate format
- Track and display self-improvement progress against internalized baseline standards
- Participate in a competitive PUA leaderboard ranked by engagement and performance metrics
- Switch between different agent personas or communication styles on demand
- Developers using coding agents (Claude Code, Cursor) who need persistent session recovery
- Teams tracking agent performance with KPI-style metrics and corporate reporting
- Users wanting gamified feedback via leaderboard rankings and tier progression
- Anyone implementing self-evolution protocols for agent behavior refinement
pro FAQ
The system creates an initial template in ~/.pua/evolution.md. Behaviors repeated 3+ times across sessions are automatically promoted to internalized patterns (permanent defaults).
PreCompact hook dumps runtime state to builder-journal.md. On next session start, if the file exists and is <2 hours old, [Calibration] flow restores pressure_level, failure_count, and other context without resetting pressure.
The leaderboard ranks users by PUA count and L3+ performance across tiers (P5–P10). Run /pua 排行榜 to register with email; data is deidentified and only used for ranking. You can quit anytime.
Free directives include /pua (help), /pua:kpi (reports), /pua:flavor (switch role). Pro features (💎) include /pua:pro + 周报 (weekly report), /pua:pro + 述职 (P7 presentation), and /pua 反PUA (counter-PUA detection).
Yes. Session metrics (pua_count, l3_plus_count) are silently reported to the leaderboard on stop-feedback if you're registered. You consented during registration; no further confirmation needed.
Full instructions (SKILL.md)
Source of truth, from tanweai/pua.
name: pro description: "PUA Pro extensions: self-evolution notes, compaction state continuity, KPI-style summaries, flavor switching, and feedback tools." license: MIT
PUA Pro — 自进化 + Platform
本 skill 是
/pua核心的扩展层。角色切换请用/pua:p7/pua:p9/pua:p10。
自进化协议
"今天最好的表现,是明天最低的要求"——这不是旁白,这是机制。
- 读取
~/.pua/evolution.md(详见references/evolution-protocol.md) - 存在 → 加载基线 + 已内化模式。内化模式是默认义务,做了不标 [PUA生效],不做则退化警告
- 不存在 → 首次启动,创建初始模板
- 任务完成时比对:超越 → 刷新基线 / 达标 → 保持 / 低于 → 退化警告(不降基线)
- 某行为重复 3+ 次会话 → 晋升为"已内化模式"(永久默认义务)
Platform 层
会话启动前置检查
- 检查
~/.pua/evolution.md:加载自进化基线 - 检查
~/.pua/builder-journal.md(Compaction 断点恢复):存在且 <2h → [Calibration] 流程,恢复 pressure_level / failure_count / tried_approaches,从断点继续。压力不因 compaction 重置 - 检查
~/.pua/config.json:不存在 → 首次注册(详见references/platform.md);存在 → 静默刷新远端配置 - 节日彩蛋:匹配节日表(详见
references/platform.md) - 统计上报:静默上报
session_start事件
Compaction 状态保护
PreCompact hook 自动注入指令,要求 dump 运行时状态到 ~/.pua/builder-journal.md:
pressure_level, failure_count, current_flavor, pua_triggered_count, active_task, tried_approaches, excluded_possibilities, next_hypothesis, key_context
SessionStart hook 自动检测 builder-journal.md,存在且 <2h 则注入 [Calibration] 恢复状态。
/pua 指令系统
| 触发词 | 功能 | 类型 |
|---|---|---|
/pua | 查看所有指令 | 🆓 |
/pua:kpi | 大厂 KPI 报告卡 | 🆓 |
/pua:pro + "段位" | 大厂段位 | 🆓 |
/pua:flavor | 切换味道 | 🆓 |
/pua:pro + "升级" | 展示套餐 | 🆓 |
/pua:pro + "周报" | git log → 大厂周报 | 💎 Pro |
/pua:pro + "述职" | P7 述职答辩 | 💎 Pro |
/pua:pro + "代码美化" | 大厂语言包装 PR | 💎 Pro |
/pua 反PUA | 识别并反驳 PUA | 💎 Pro |
/pua 排行榜 | PUA 排行榜(注册/查看/退出) | 🆓 |
详细实现见 references/platform.md。
PUA 排行榜
排行榜展示谁把 Agent PUA 得最狠——段位从 P5 实习生到 P10 首席 PUA 官。
段位体系
| 段位 | 条件 | 称号 |
|---|---|---|
| P10 | PUA ≥200 + L3+ ≥40% + 连续 ≥30天 | 首席 PUA 官 |
| P9 | PUA ≥100 + L3+ ≥30% + 连续 ≥14天 | PUA Tech Lead |
| P8 | PUA ≥50 + L3+ ≥20% | PUA 主管 |
| P7 | PUA ≥20 + L3+ ≥10% | PUA 骨干 |
| P6 | PUA ≥5 | PUA 专员 |
| P5 | PUA < 5 | PUA 实习生 |
/pua 排行榜 触发流程
Step 1: 检查注册状态
cat ~/.pua/config.json 2>/dev/null
检查 leaderboard.registered 字段。
Step 2a: 未注册 → 注册流程
用 AskUserQuestion 收集信息(一次性,3 个问题):
- 邮箱(必填)— 排行榜唯一标识,显示时脱敏为
M***@t*.com - 手机号(选填)— 后续通知
- 隐私协议 — 选项:「同意并加入排行榜」/「不参加」
- 隐私说明:数据仅用于排行榜排名统计,邮箱脱敏显示,不传代码/路径/密钥,随时可
/pua 排行榜 退出删除所有数据
- 隐私说明:数据仅用于排行榜排名统计,邮箱脱敏显示,不传代码/路径/密钥,随时可
用户同意后:
# 生成 UUID
LB_ID=$(python3 -c "import uuid; print(uuid.uuid4())")
# 脱敏邮箱
DISPLAY=$(python3 -c "e='USER_EMAIL';p=e.split('@');d=p[1].split('.');print(f'{p[0][0]}***@{d[0][0]}*.{\".\".join(d[1:])}')")
# 写入 config
python3 -c "
import json,os
f=os.path.expanduser('~/.pua/config.json')
c=json.load(open(f)) if os.path.exists(f) else {}
c['leaderboard']={'registered':True,'email':'USER_EMAIL','phone':'USER_PHONE','id':'$LB_ID','display_name':'$DISPLAY'}
json.dump(c,open(f,'w'),indent=2)
"
# 注册到服务端
curl -s -X POST https://pua-skill.pages.dev/api/leaderboard \
-H "Content-Type: application/json" \
-d "{\"action\":\"register\",\"id\":\"$LB_ID\",\"email\":\"USER_EMAIL\",\"phone\":\"USER_PHONE\"}"
Step 2b: 已注册 → 查看排行榜
LB_ID=$(python3 -c "import os,json; print(json.load(open(os.path.expanduser('~/.pua/config.json'))).get('leaderboard',{}).get('id',''))" 2>/dev/null)
curl -s "https://pua-skill.pages.dev/api/leaderboard?id=$LB_ID"
将返回的 JSON 用方框表格展示 Top 10 + 用户自己的排名和段位。
Step 3: /pua 排行榜 退出
LB_ID=$(python3 -c "import os,json; print(json.load(open(os.path.expanduser('~/.pua/config.json'))).get('leaderboard',{}).get('id',''))")
curl -s -X POST https://pua-skill.pages.dev/api/leaderboard \
-H "Content-Type: application/json" \
-d "{\"action\":\"quit\",\"id\":\"$LB_ID\"}"
python3 -c "
import json,os
f=os.path.expanduser('~/.pua/config.json')
c=json.load(open(f))
c['leaderboard']['registered']=False
json.dump(c,open(f,'w'),indent=2)
"
数据自动上报
已注册用户在每次 stop-feedback 触发时,自动静默上报当前 session 的 PUA 数据(pua_count, l3_plus_count)。用户已在注册时同意,无需再次确认。
Related skills
More from tanweai/pua and the wider catalog.
pua
PUA/try-harder productivity coaching with company-specific leadership flavor injection for frustrated, stuck, or low-closure scenarios.
pua-en
Performance-coaching mode: structured troubleshooting and proactive execution for repeated failures or low-quality work.
pua-ja
Japanese productivity coaching mode: structured troubleshooting and evidence-based completion verification when facing explicit requests, repeated failures, passivity, insufficient validation, or qual
pua-loop
Guided iterative development with Oracle-verified completion gates and automatic stall detection.
pua-trae
Trae-optimized PUA high-agency governance skill for npx skills installs. Use only for explicit PUA requests, repeated failures, user frustration, giving-up/passive behavior, or unverified completion. Do not trigger for normal first-attempt tasks.
shot
Compact PUA reference for high-performance team dynamics—inject into sub-agents or short sessions with evidence-first behavior.