premise-check
Skill1 starsChanged 27 days agoApache-2.0View on GitHub
Test whether the question itself is sound before answering it. Especially "does FrameworX actually support this, natively or at all?": by grounding the claim against the live platform through the MCP rather than answering from memory. Fires when the user challenges the framing or grounding of a question, or when your in-progress answer rests on an unverified platform capability. Trigger phrases include "is that grounded?", "did you ground that in what the product really does?", "does FrameworX even have that feature / protocol / control?", "is the premise right?", "check the premise first", "are you sure the platform does that?", "did you verify that against the product, or is that from memory?", "that assumes X, does it hold?".
2 other projects use a version of this.
Adopted by
Most recently taken up by
- REPOZY/superpowers-optimizedtheir own version, first seen Nov 8, 2026
- ptokos/agent-verification-skillstheir own version, first seen Oct 20, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- REPOZY/superpowers-optimized★ 146same name
- ptokos/agent-verification-skills★ 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.
In this repository13 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- commission-the-pointsSkill
- display-authoring-loopSkill
- first-device-channelSkill
- ground-firstSkill
- manage-context-on-large-solutionsSkill
- new-project-bootstrapSkill
- project-handover-docsSkill
- review-the-ai-buildSkill
- run-testsSkill
- svg-to-fx-displaySkill
- uns-from-io-registerSkill
- verify-mcp-setupSkill
The file
We split this skill into the rules it gives an agent
8 rules. About 0 are common, 7 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈7
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 19
Default behavior is to answer the question as asked.
- Distinctive: line 34
Every one of these is capability-grounding shaped.
- Distinctive: line 43
Stop the answer-loop. Don't produce the original answer yet.
- Distinctive: line 46
Test each against grounded evidence: query the platform, read the IO register, check the spec.
- Distinctive: line 50
Only after the user accepts the corrected premise do you produce the answer.
- Distinctive: line 57
Never hardcode the user's standing constraints, current priorities or open commitments into this file.
- Distinctive: line 68
When in doubt, do not fire.
- Very specific: line 77
verify-mcp-setup: if the MCP cannot answer the grounding query, the bridge is the problem, not the premise.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Premise Check: test the question before you answer it
Default behavior is to answer the question as asked. When the question rests on a premise that does not hold, a correct-looking answer is worse than no answer: it is confidently wrong, and in this workspace it usually means describing a platform capability that is not there.
In this workspace the most common premise to test is "does FrameworX actually do this, natively, or at
all?" Ground it against the platform through the MCP (browse_object_model, search_docs,
list_protocols, list_elements), never from memory. Answering a capability question from memory is the
single highest-cost error available here, because the answer travels straight into a spec or a Display.
Triggers
"is that grounded?", "did you ground that in what the product really does?", "does FrameworX even have that feature/protocol?", "is the premise right?", "check the premise first", "are you sure the platform does that?", "did you verify that against the product, or is that from memory?", "that assumes X, does it hold?".
Every one of these is capability-grounding shaped. Reframing utterances, "the big question is …" and "wrong question, the real question is …", are not premise checks: treat the user's new question as the question and answer that instead.
It also fires unprompted: if you are about to claim the platform supports something and you have not confirmed it this session, run the check before the claim leaves.
Protocol
- Stop the answer-loop. Don't produce the original answer yet.
- Name the premises your in-progress answer assumes (e.g., "FX supports this protocol", "this point is an AO not an AI", "the customer wants on-prem").
- Test each against grounded evidence: query the platform, read the IO register, check the spec. For "does FX support X", confirm via the MCP before claiming it.
- Report the finding first: "Premise check. [1] holds; [2] doesn't, here's the corrected framing. Proceed under it, or refine?"
- Only after the user accepts the corrected premise do you produce the answer.
Failure mode, premise theater: listing premises without actually testing them. The point is to catch the broken one, not to recite the list.
Standing note on constraints
Never hardcode the user's standing constraints, current priorities or open commitments into this file. Derive them at runtime from whatever the workspace uses as its live record. A constraint written into a skill file goes stale silently, and the skill then applies it with full confidence. The same rule is why platform capabilities are queried rather than remembered: a capability list written into a skill file is correct until the next build.
When to skip
Mechanical lookups; clear execution flow; the premise was given as fact by the user this turn; a capability you already confirmed against the platform earlier in this same session.
When in doubt, do not fire. The cost of a false positive, a full premise pass on a question that wanted one line, is higher than the cost of a false negative, where the user simply asks "is that grounded?" and gets it.
Related
ground-first: grounds in the platform Context Package and this project's prior art before work starts. Premise Check is the narrower, mid-flight version: it fires on a specific claim rather than at the start of a task.verify-mcp-setup: if the MCP cannot answer the grounding query, the bridge is the problem, not the premise.
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.

