notwitcheer/hermes-recipes/daily-brief
Claude Code and agent skill by notwitcheer/hermes-recipes
- community
Checking that the SKILL.md file is still in the repository.
Compose the operator's morning brief: stack health + a dated news digest + a kickoff nudge. Triggered by the morning cron. OUTPUT: terse — bullet points, minimal prose, scannable. Under ~25 lines.
Install
npx vskill@latest install notwitcheer/hermes-recipes/daily-briefSKILL.md
Daily Brief
When to Use
Triggered by the morning cron job (hermes cron add — see the recipe README). Produce ONE concise message with three sections, in the agent's voice. Keep the whole thing under ~35 lines.
Procedure
- Stack health (local, via terminal tool — run these and summarize in 3-4 lines):
nvidia-smi --query-gpu=temperature.gpu,utilization.gpu,memory.used,memory.total --format=csv,noheadersystemctl is-active llama-server; systemctl is-active hermes-gatewaydf -h / | tail -1grep -iE "error|traceback|exception" ~/.hermes/logs/*.log 2>/dev/null | tail -5(say "clean" if none)
Flag anything red: a service not "active", disk >85%, GPU temp >80C.
- News digest (local script — every candidate carries a machine-verified date):
run ~/.hermes/scripts/news_digest.sh via the terminal tool. It prints dated candidates from configured sources (community threads ranked by real vote counts, fresh GitHub releases). Pick 2-3 genuinely interesting items FROM THAT LIST ONLY. For EACH item:
- write ONE complete sentence, grounded only in what the candidate line actually
says — no invented numbers or claims;
- KEEP THE ITEM'S [date] in your line, e.g. "(Jul 7)" — never say "just released"
or "new" for anything older than yesterday; the date decides, not you;
- end the line with the item's real
https://...source URL.
You may web_extract a chosen item's URL to enrich the sentence, but never to add new items. If the script prints "no dated candidates", write "digest: quiet night" and continue — never block the brief, never fabricate sources.
- Kickoff nudge: one short line — what's worth building today, or a question.
Why the digest is a script and not web search
An agent-driven web search returns undated snippets, and the model will narrate anything as fresh news — a month-old release reads exactly like yesterday's. The architecture rule this skill encodes: the model WRITES, it never researches. Machine-dated candidates come from a deterministic read-only script; the model's job is selection and one good sentence per item.
Pitfalls
- The high GPU utilization you see during this run is YOUR OWN inference generating
this brief — never flag it as a runaway process.
- This runs unattended; if a command errors, note it briefly and move on rather than
aborting.
- After
timedatectl set-timezone, restart the gateway or this cron fires on the old
timezone.
Security
- Security scanPassed, 100 of 100Pattern checks on the skill's files.
- Model reviewNot runA model review is scheduled for the next rescan.
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingrecipes/sovereign-gpu/skills/daily-brief/SKILL.md
Popularity
- Repository stars
- 138
- Repository forks
- 10
- 7-day trend score
- 30
Stars and forks are for the whole GitHub repository, not this skill alone. Not refreshed yet.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/notwitcheer/hermes-recipes/daily-brief)