-
Notifications
You must be signed in to change notification settings - Fork 2.9k
[Resource]: cstack #1363
Description
Display Name
cstack
Category
Workflows & Knowledge Guides
Sub-Category
None
Primary Link
https://github.com/srf6413/cstack
Author Name
srf6413
Author Link
License
No License / Not Specified
Other License
No response
Description
A pattern for building autonomous Claude agents without infrastructure: each domain gets a Notion state page (SSOT), a SKILL.md behavioral contract, a Cowork cron schedule, and a notification channel. Agents are stateless at runtime — read state at session start, do work, write updates at session end. The human is in the loop by design, called only when a decision requires judgment; no code, no servers, no always-on processes.
Validate Claims
The README contains a 5-minute setup prompt. Paste it into a Claude Cowork session, answer 6 questions about your domain, and it generates 3 assets: a Notion state page, a SKILL.md behavioral file, and a session schedule. Run the scheduled session twice. The second run will read what the first run wrote and continue from that state. No setup beyond a free Notion account and a Cowork subscription.
Specific Task(s)
Run the 5-minute setup prompt from the README to create an assistant for any recurring task (e.g., "track my open reading list and surface the 3 highest-priority items each morning"). Schedule it as a Cowork session. Run it once manually, then let it run on schedule. Check the Notion state page after each run to see state accumulate.
Specific Prompt(s)
Copy the prompt block under "Try it in 5 minutes" in the README and paste it into a Claude Cowork session. Answer the 6 questions. You'll receive a complete Notion page template, a SKILL.md file, and scheduling instructions — ready to deploy with no modifications.
Additional Comments
cstack is a pattern, not a library. The value is the design: stateless sessions + external persistent memory + behavioral contracts + human-in-the-loop escalation. Everything needed to replicate it is in the repo. The README covers both non-technical setup (5-minute wizard) and architecture (Technical Appendix section with stateless design, externalized cognition thesis, and parallel execution model).
Recommendation Checklist
- I have checked that this resource hasn't already been submitted
- It has been over one week since the first public commit to the repo I am recommending
- All provided links are working and publicly accessible
- I do NOT have any other open issues in this repository
- I am primarily composed of human-y stuff and not electrical circuits