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 agentmods add skills/channinglua/prax-agent/support-digestnpx skills add ChanningLua/prax-agent --skill support-digestgit clone --depth 1 https://github.com/ChanningLua/prax-agentWhat 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 | $0.00031 | $0.02127 |
| Opus 5 | $0.00015 | $0.01064 |
| Sonnet 5 | $0.00006 | $0.00425 |
| Haiku 4.5 | $0.00003 | $0.00213 |
Grade A, and why
support-digest scanned grade A with 0 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 2d 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 211 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Support Ticket Digest
痛点：PM 每天早上要花 20 分钟翻昨天 200 条客服工单，才能知道"哪个 feature 坑多"、"有没有新涌出的抱怨"。这个 skill 把翻 ticket 的动作自动化，所有数据留在本地，不调任何外部 API，方便合规严格的团队。
何时触发
- cron 每天早上 9 点跑
- 用户手动："生成昨日客服简报"
- 客服系统 CSV/JSON 导出到指定目录触发 hook
输入
- 必需：
.prax/inbox/tickets-<YYYY-MM-DD>.json（或.csv也支持）- 标准字段：
id / created_at / status / category / subject / body / customer_email / severity - 字段缺失时尽量宽容，但
id / created_at / body必须有
- 标准字段：
- 可选：
.prax/support-digest.yaml配置（见下方）
Prax 不负责从 Zendesk / Freshdesk 拉数据——export 是 IT 或客服负责人的事。skill 只做本地文件处理。
输出
.prax/vault/support/<YYYY-MM-DD>/digest.md
外加原始统计数据的结构化副本：
.prax/vault/support/<YYYY-MM-DD>/stats.json
用于下游工具（BI、周报生成等）复用。
工作流程
Step 1：定位输入
# 最新的 tickets 文件
ls -t .prax/inbox/tickets-*.{json,csv} 2>/dev/null | head -1
没找到 → 停下来告诉用户："请把昨日工单导出为 JSON/CSV 放到 .prax/inbox/tickets-YYYY-MM-DD.json"。
Step 2：加载 + 脱敏
读进来后立刻脱敏用户信息：
| 字段 | 脱敏规则 |
|---|---|
customer_email |
[email protected] → f***@b***.com |
customer_phone |
13812345678 → 138****5678 |
| body 里的 email | 正则替换同上 |
| body 里的手机/银行卡号 | 正则识别，全替换成 [REDACTED] |
| body 里出现的 SSN / 身份证号 | 全替换成 [REDACTED] |
脱敏后才进入后续处理——避免不小心把 PII 写进 digest。
Step 3：分类统计
按 category 字段（或从 subject 启发式推断）聚类：
- 计数：今日 vs 昨日 vs 过去 7 天均值
- 热度：新增占比（category_today / category_7d_avg）
- 严重程度：
severity=high的占比 - 新类别：今日有而过去 7 天没出现过的 category
Step 4：识别 top issues
抽取前 5 条"关注度高"的工单，标准（按分数降序）：
| 信号 | 分数 |
|---|---|
| severity=high | +3 |
| status=escalated / status=re-opened | +2 |
| 同一 category 今日数量 > 7d_avg × 2 | +2 |
| body 包含 "refund" / "退款" / "退费" | +2 |
| body 包含 "lawyer" / "律师" / "投诉" / "举报" | +3 |
| response_time > 24h（如果有此字段） | +1 |
取前 5 条，每条附脱敏后的 body 摘录（≤ 100 字）。
Step 5：趋势识别
对比"今日 vs 过去 7 天均值"：
- 📈 涨幅 > 50% 的 category：列出来
- 📉 跌幅 > 50% 的 category：列出来（有时代表问题解决了）
- 🆕 今日首次出现：全列
Step 6：写 digest
模板：
---
date: 2026-04-22
ticket_count: 187
category_count: 12
highlights_count: 5
generated_at: 2026-04-22T09:05:00+08:00
---
# 客服简报 · 2026-04-22
## 总览（一句话）
昨日共 **187** 条工单，比 7 日均值 +12%；`billing` 类翻倍，`auth` 类回落。
## 今日亮点（top 5）
### 1. [high] Billing — duplicate charge
- 工单数：23（占 12%）
- 代表性摘要："On 4-20 I was charged twice for the Pro plan. Please refund immediately."
- 建议 owner：@finance-ops
### 2. [escalated] 登录流 — OAuth timeout
- ...
## 趋势
### 📈 涨
- `billing`：23 vs 7d 均值 11（+109%）
- `mobile-app-crash`：8 vs 3（+167%）
### 📉 跌
- `auth`：12 vs 19（-37%）
### 🆕 新出现
- `integration-slack`（6 条）
## 分类统计
| Category | Today | 7d avg | Δ |
|---|---|---|---|
| billing | 23 | 11 | +12 |
| auth | 12 | 19 | -7 |
| ...|
## 报告位置
- Raw 脱敏 JSON：`.prax/vault/support/2026-04-22/tickets-redacted.json`
- Stats：`.prax/vault/support/2026-04-22/stats.json`
- 原始文件已归档到：`.prax/inbox/archive/tickets-2026-04-22.json`
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.
- 2d ago First seen · 211 lines · 31 tokens per session scan A a4855375207c
support-digest is a skill published in the GitHub repository ChanningLua/prax-agent (272 stars, last pushed 1mo ago), licensed MIT. It adds 31 tokens to every session and 2,127 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
start9-frontend
Start9 house style for every Angular + Taiga UI web frontend (StartOS ui/setup-wizard, StartWRT, StartTunnel, brochure-marketplace, start9-store, ops-server, support-server). Use BEFORE writing or reviewing any frontend code in these apps — components, templates, styles/CSS, forms, dialogs/toasts, routing, state, DI…
designer_guide
Guidance for creating and editing designs using the designer tool.
agents_guide
Guidance for delegating focused work to child Agents.
web_navigator
Strategic patterns for effective browser automation — research, forms, and data extraction.
data_analyst
Analyses datasets with professional rigour — statistical summaries, clear narratives, and well-chosen visualisations.
humanizer
Write in a natural, human tone — no AI-speak, no filler, no corporate fluff.