Missing layer: /avl — Assumption Verification before planning
- Dominant language
- TypeScript
- Stars
- 133k
- Forks
- 19.9k
- Avg merge
- 18h 46m
- Merged PRs (30d)
- 26
Description
Hi Garry,
I've been studying gstack closely and found a gap in the workflow.
Every skill in gstack assumes the input is correct. The CEO review challenges feature framing. Engineering manager locks architecture. Designer rates decisions. QA tests what was built.
None of them ask: "Is the assumption underneath this decision actually true?"
That's where most product failures live — not bad code, not bad design — but an unverified assumption nobody questioned before the sprint started.
I built /avl — an Assumption Verification Layer skill that sits before /office-hours in the workflow:
Input → /avl → verify assumption → signal check → /office-hours → /autoplan → build
The skill:
1. Surfaces the core assumption underneath the decision
2. Classifies it (market/technical/resource/timing/problem)
3. Asks one precise verification question
4. Checks if signal is verified or unverified
5. Either proceeds to planning or gives a 48-hour validation test
Free. MIT license. One file.
Repo: https://github.com/01382253386/avl-skill
This is based on Selfune AI — a cognitive interrupt layer I've been building for human decision-making, backed by a 1,339-row dataset across 39 bias categories.
Would love to see /avl added to the gstack workflow between Think and Plan.
— Saurabh Dhuliya, Selfune AI
Contributor guide
Assessment
This issue has not been assessed yet.