Separate facts from assumptions
Mark what is established, what is open and what is only a possible blind spot.
Skill
Unknowns
Find the question that could change your plan.
For people starting unfamiliar or uncertain work. Unknowns makes assumptions and unanswered questions visible, attaches an evidence basis and helps choose the next useful inquiry instead of growing an endless risk list.
Established: the consumer accepts the current export Open question: does it accept the proposed schema revision? Unstated preference: is manual conversion acceptable to the owner? Possible blind spot: a downstream consumer may rely on an undocumented field Next inquiry: test the proposed export against the named consumer Evidence needed: actual parse result and supported schema contract
Fictional migration map. The possible downstream dependency is a hypothesis, not a discovered dependency.
Spend inquiry where it matters
Choose a question whose answer could alter a decision or acceptance test.
Keep a map that can change
Reclassify an entry when evidence arrives, retaining the source and the reason for the change.
The film
Twelve green steps. One secret.
In 67 seconds, a migration plan looks perfect until step seven turns out to rest on a guess nobody checked. Unknowns sorts what you know from what you are only assuming, picks the one question that could change the plan, and turns it into a small test before anything moves. Captions are on the player; a transcript is below.
Read the transcript
This plan is feeling great. Twelve steps. All green.
But step seven has a secret. Nobody checked it.
It assumes the new system can read last year’s files.
Rewind. Unknowns sorts what you know from what you’re only guessing.
Then it shines a light on the one question that could change everything.
Can it read last year’s files? Open twenty, before you move anything.
Eighteen open. Two don’t. Better to learn that now.
When the answer comes back, the map changes, and it keeps the source.
On screen: Test the one thing that could sink the plan.
Unknowns. Find the question that could change your plan.
The migration plan, the files and the test result are a fictional example. The film shows how the skill behaves; it does not report results.
See the work take shape.
Fictional migration map. The possible downstream dependency is a hypothesis, not a discovered dependency.
Read the actual territory
Use the code, records and user constraints as evidence rather than filling quadrants with guesses.
Name the consequential gap
Record its evidence basis, severity and the decision it could change.
Choose and update
Select one inquiry, gather the evidence and update only the entries it resolves.
Fictional example
Test an export assumption
The brief Use Unknowns to review a fictional export migration plan. Separate established requirements from compatibility assumptions, attach an evidence basis and choose the one unanswered question most likely to change the plan. Keep possible blind spots labeled as hypotheses.
Before
A migration plan treats an untested compatibility assumption as settled.
After applying the method
The map records it as a consequential question with an explicit verification step.
The next inquiry targets a decision-changing unknown.
A clear job.
A useful stopping point.
Find blind spots, map uncertainty, or decide what to investigate next.
Use it when
- A plan rests on an untested compatibility or customer assumption.
- You are entering a codebase or domain you do not know well.
- A team needs to choose what to investigate next.
Choose another route when
- Claiming a brainstormed hazard has been discovered as a fact.
- Filling every quadrant just to make the diagram look complete.
Map
A plan needs an uncertainty inventory.
An evidenced four-quadrant map with priorities.
Blindspot
The user wants to surface unstated assumptions.
A conversational challenge and concrete test.
Resolve
An important uncertainty is actionable.
A selected inquiry, evidence and a reclassified map entry.
Read the canonical method
Unknowns
Map the task against the actual territory: code, records and user constraints. Separate established claims (known knowns), explicit questions (known unknowns), unstated assumptions (unknown knowns) and possible blind spots (unknown unknowns). Do not force every quadrant to contain an item.
Map: read references/unknowns-matrix.md and references/unknownfinder.md. Attach an evidence basis and severity to each consequential item. Possibilities are hypotheses, not discovered facts. Prefer a useful question over a large inventory of speculative hazards.
Blindspot: stay in the current conversation. Surface one assumption or ask one question whose answer could change the plan. Do not delegate away the interaction when the user's answers are the needed evidence.
Resolve: use the retained unknowns CLI and unknowns_map library for deterministic map operations. Seed, add, reclassify, status and triage work without an engine or API key. Keep the canonical machine DOT separate from the pretty renderer. Graphviz is optional for PNG export. Select the next inquiry from the dominant unresolved quadrant and update its status only when evidence changes.
The existing /unknownfinder alias and agents remain optional Amplifier adapters to the same method. Pipeline execution is optional and requires its own configured engine; a rendered map or routing token is not evidence that the inquiry ran. Generated .ai/unknowns-map.dot belongs to the user's task, never to the source release.
Limits and prerequisites
- A brainstormed blind spot is a hypothesis; classification does not prove it exists.
Your first useful run.
Bring one plan and ask which unanswered question would most change its next step.
The download contains the complete skill folder. Extract it into the skills directory supported by your agent, keeping its references and license together. Or use the source installer below. Python 3.11+ is required for that installer.
Start from the public source
git clone https://github.com/michaeljabbour/amplifier-skill-unknowns.git
cd amplifier-skill-unknownspython3 scripts/capability.py install --dest ~/.agents/skillsNative tools and optional execution
python3 -m pip install .These commands are for the optional native surface. Follow its installation guide and prerequisites; portable method installation is separate.
Keep the complete skill folder together, including its references and license. The host applies the method in your context; the skill does not start a background service.
- Python 3.11 or later is required by the portable install/build helper. The installed method runs through your agent host.
- The portable method and deterministic map operations need no engine or model key.
- The optional native unknowns CLI can seed, add and triage a map; Graphviz is only needed for image export.
- Pipeline execution is separate from map creation and needs its own configured runner.
Try this with your agent
Use Unknowns to review a fictional export migration plan. Separate established requirements from compatibility assumptions, attach an evidence basis and choose the one unanswered question most likely to change the plan. Keep possible blind spots labeled as hypotheses.Use after installing in a host that loads skill folders. Supply your own task files and review the result.
Keep the guide nearby
Read the guide · License · Changes · Release metadata · CONSOLIDATION
Know what you can rely on.
Native map/tool tests and an installed no-engine seed/add/triage check passed. A map or routing token does not demonstrate that an inquiry ran or that rework was avoided.
Read the checks and their scope
- Functional acceptance — passed
31 native map/tool tests passed; installed no-engine seed/add/triage gate also passed.
- Native host acceptance — not-run
No live model, real desktop transaction, invitation, deployment or production data migration was performed by this functional baseline.
Source and conformance checks are separate from demonstrated benefit. Comparative effectiveness and outside-user outcomes are not established by this site.
Fictional examples explain the capability. Tests check the stated behavior. Outside-user outcomes need their own evidence.