finding-unknowns
Skill1 starsChanged 3 months agoApache-2.0View on GitHub
Use whenever requirements are ambiguous, vague, or underspecified — no matter how small the task sounds ("make it look better", "clean this up", "nothing fancy") — and when starting any non-trivial task: unfamiliar domain or codebase, client-facing deliverables, or changes spanning multiple files. Run before writing plans or code; again during implementation when edge cases force deviation from the plan; and after implementation before handoff, sign-off, or merge. Especially when time pressure or an unreachable user tempts you to skip clarification and build on assumptions.
4 other projects use a version of this.
Adopted by
Most recently taken up by
- droppedoutofcontext/finding-unknownstheir own version, first seen Nov 1, 2026
- JPeetz/agent-skillstheir own version, first seen Oct 28, 2026
- HarryJhin/groundworktheir own version, first seen Aug 4, 2026
- sidneyswift/skillstheir own version, first seen Jul 21, 2026
Other versions4 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 4
- JPeetz/agent-skills★ 14same name
- sidneyswift/skills★ 9same name
- droppedoutofcontext/finding-unknowns★ 0same name
- HarryJhin/groundwork★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
The file
We split this skill into the rules it gives an agent
3 rules. 0 are common, 3 are distinctive, and 0 are very specific to this project.
0
Common
Most of the 3 edited versions we read keep these.
3
Distinctive
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 29
Before writing any plan or implementation code, your FIRST output is an Unknowns Inventory:
- Distinctive: line 38
Every item gets a resolution route: ask the user, request a reference, build a cheap prototype/mockup, or — only when the user is …
- Distinctive: line 57
When the plan gets written, resolved items feed it as requirements.
“Common” is measured: kept by at least 6 in 10 of the 3 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Finding Unknowns
Overview
The request is a map; the codebase and real-world constraints are the territory. Every gap between them gets filled with a guess — cheap to surface before implementation, expensive to discover after delivery.
Core principle: spend minutes surfacing unknowns before spending hours building on them.
The Four Quadrants
| Quadrant | What it is | How to eliminate |
| Known knowns | Stated in the request | Already on the map |
| Known unknowns | Questions noticed but unanswered | Interview: ask, one at a time |
| Unknown knowns | Obvious to the requester, never stated — house templates, client standards, contract formats | Prototype/mockup (seeing it exposes the standard); ask for a reference example |
| Unknown unknowns | Questions nobody thought to ask | Blind spot pass |
The Rule
Before writing any plan or implementation code, your FIRST output is an Unknowns Inventory:
## Unknowns Inventory
- [known unknown] <question> → resolve by: ask / reference / conservative default (logged)
- [unknown known] <suspected unstated standard> → resolve by: mockup / example / ask
- [blind spot] <area not yet examined> → resolve by: blind spot passEvery item gets a resolution route: ask the user, request a reference, build a cheap prototype/mockup, or — only when the user is unreachable — take the conservative default and log it as a deviation.
An unreachable user makes the inventory MORE important, not less: it decides which defaults are safe and becomes the assumption log delivered with the work.
Scale the inventory to the ambiguity, not the ritual. When the request names its exact targets and leaves no format, convention, or scope choice open (a rename, a typo fix, a version bump), the whole inventory is one verification action — confirm the stated scope (e.g. a repo-wide grep), then do the work. The printed block earns its length only when items genuinely route to different resolutions.
Phase Routing
| You are... | Read |
| Starting a task (no code written yet) | references/pre-implementation.md |
| Mid-implementation, plan didn't cover this case | references/during-implementation.md |
| Done, preparing handoff / review / sign-off | references/post-implementation.md |
From Inventory to Plan
The inventory is not a report — it drives what happens next:
- It is the question list for whatever design discussion follows; work through it with the user before committing to a direction.
- When the plan gets written, resolved items feed it as requirements; unresolved items enter it as documented assumptions, never as silent guesses.
Rationalizations
| Excuse | Reality |
| "User is offline and the deadline is now" | The inventory takes two minutes and changes what you build. Do it, pick conservative defaults, log every one. |
| "Waiting for clarification costs more than delivering on assumptions" | Unlogged assumptions cost the most — the user finds them after delivery. Speed and surfacing unknowns are not in conflict. |
| "The request is clear enough" | Clear requests still hide the requester's unstated standards. One mockup exposes them. |
| "I'll list my assumptions at the end" | By then the work is already shaped by them. Surface them first. |
| "Asking questions looks incompetent" | Shipping the wrong thing looks worse. |
Red Flags — STOP and build the inventory
- About to write code for a one-sentence task description
- Choosing a format, library, or convention the user never mentioned
- Thinking "they probably want..."
- A client-facing or sign-off deliverable with no example of what the recipient expects
- Three unstated decisions made in a row
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.