convez.ai
SkillAutomaçãoAvançado
encircleacity2

Slack Daily Digest para Feishu

encircleacity2/slack-daily-digest

Sintetiza Slack diariamente em resumos estruturados no Feishu com IA

Ver no GitHub
InicialPluginScore · de 0 a 100

Skill que transforma seu Slack em um resumo inteligente diário sintetizado por IA no Feishu (Lark), cobrindo destaques, TODOs, pipeline e riscos. Executa duas vezes por dia, fixa o resumo em um documento índice unificado e notifica via Lark IM com watchdog auto-curativo.

Instalação

$npx convez-ai encircleacity2/slack-daily-digest

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
0
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 6 meses
Versão
v1.0.0
Issues abertas
0
Contribuidores
1

Casos de uso

  • Executivos acompanharem resumo diário de atividades por canal
  • Times de produto rastrearem highlights, risks e pipeline em um documento único
  • Automação de relatórios diários sem intervenção manual

Veredito

Skill bem-estruturado com orquestração sofisticada (watchdog, retry backoff, synthesis periódica). Requer setup complexo (Slack App + Lark Bot) e dependência de skills Mira adjacentes, adequado para times que já usam Slack + Feishu. Implementação técnica sólida com tratamento de falhas.

README

A tradução para o português deste README ainda está na fila.

slack-daily-digest

A Mira / Claude Agent skill that turns your Slack workspace into an AI-synthesized daily intelligence digest in Feishu (Lark) docs — Highlights, TODO backlog, business pipeline, risks, by-channel breakdown — pinned to a single index doc and pushed to you via Lark IM.


✨ Features

  • One-time onboarding walks you through creating a Slack User OAuth App and a Lark Custom Bot, then stores credentials locally (~/.slack-daily-digest/credentials.json, mode 0600).
  • Unified index doc in Feishu — every daily digest is pinned to the top of the list.
  • Twice-daily synthesis (default 11:00 / 16:00 UTC+8, user-configurable) covering recent messages with:
    • 🎯 Executive Summary (P0/P1/P2 prioritized)
    • 📌 Highlights · ✅ TODO · 📈 Pipeline · ⚠️ Risk Signals
    • 🗂️ By-Channel verbatim transcripts (no truncation, no ellipsis)
    • 📝 Statistics + identified BytePlus / ByteDance owners
  • Self-healing watchdog runs T+20min after each scheduled slot. Verifies all required sections exist; on failure self-retries with [60s, 300s, 900s] backoff up to 3 times, then escalates.
  • Lark IM card notification with Highlights + TODO summary + "Open Full Digest" button.

📦 Install

This is a Mira / Claude-Agent skill. Drop the contents of this repo into your skills directory:

git clone https://github.com/<your-org>/slack-daily-digest.git \
  ~/.claude/skills/slack-daily-digest

Then in your next conversation, ask the agent:

"Set up Slack daily digest"

The skill will guide you through Slack App + Lark Bot onboarding.


🔧 Prerequisites

RequirementWhy
Python 3.9+Helper scripts
curl, jqAPI calls + JSON parsing
A Slack workspace where you can install a custom appRead channel history
A Feishu / Lark tenant where you can publish a custom appWrite docs + send IM
The lark-docs-skill and lark-drive-wiki-skill Mira skills (or equivalent CLIs)Doc CRUD + folder ops

See references/slack-app-setup.md and references/lark-bot-setup.md for credential creation.


🚀 Quick Start

After installing, the skill auto-runs Phase A (Onboarding) on first invocation:

  1. Slack: create app → add User Token Scopes → install → paste xoxp-...
  2. Lark: create app → add scopes → publish → paste App ID + App Secret + your work email
  3. Index doc: pick a parent folder (or create new) → confirm daily run time
  4. The skill stores everything in ~/.slack-daily-digest/ and registers a recurring task

From then on, every scheduled slot:

slack-daily-digest run --slot 11:00
slack-daily-digest watchdog --slot 11:00   # T+20min

🗂️ Repo Layout

SKILL.md                          # Skill manifest (Mira-compatible)
scripts/
├── save_credentials.py           # Persist Slack/Lark creds + schedule
├── refresh_users_cache.py        # Refresh users.list cache (Tier 2, 7-day TTL)
├── list_channels.py              # users.conversations paginated list
├── pull_history.py               # conversations.history per channel + filters
├── build_digest.py               # Resolve mentions, BytePlus owner detection
├── lark_helpers.py               # tenant_token / email→open_id / doc CRUD
├── run_digest.py                 # Orchestrator (Steps 0-4 + agent handoff)
├── watchdog.py                   # T+20min self-healing checker
└── notify_lark.py                # IM card with Highlights + TODO
references/
├── slack-app-setup.md            # Step-by-step Slack App guide
├── lark-bot-setup.md             # Step-by-step Lark Bot guide
└── digest-template.md            # Synthesis template v4 (Hard Rules + rubric)

🔐 Security

  • Credentials live only in ~/.slack-daily-digest/credentials.json with mode 0600.
  • The Lark tenant_access_token is cached for 90 minutes in ~/.slack-daily-digest/.lark_tenant_token.
  • The skill never sends credentials over the wire to anything other than the Slack and Lark APIs.
  • .gitignore excludes all runtime data — verify before committing.

If your Slack token is revoked or rotated, re-run:

python3 scripts/save_credentials.py --slack-token "xoxp-..."

⚠️ Hard Rules (encoded in references/digest-template.md)

  • ❌ Never truncate a message with [:N].
  • ❌ Never collapse long messages into ellipsis.
  • ❌ Never keep only a summary while dropping the original text.
  • ✅ Filter channel_join, channel_leave, pure-emoji replies, pure greetings.
  • ✅ Allowed to merge consecutive messages from the same speaker for compactness.

🤝 Contributing

PRs welcome. Please:

  1. Strip any real credentials, doc URLs, channel IDs, or user names from logs/screenshots.
  2. Test the onboarding flow end-to-end on a fresh ~/.slack-daily-digest/ directory.
  3. Update CHANGELOG.md with user-visible changes.

📜 License

MIT

Ver o original no GitHub

Bom para

Relacionados