loft
Skill4 starsChanged todayMITView on GitHub
[oh-my-agy] Shape-before-steel prototyping: settle a design or logic fork that prose cannot resolve by building a throwaway artifact (pure zero-I/O logic module in an HTML harness, or 3 structurally distinct UI variants behind one switcher). Use for $loft or when a spec/interview stalls on 'we need to see/feel it first'.
2 other projects use a version of this.
Adopted by
Most recently taken up by
- Yeachan-Heo/oh-my-claudecodetheir own version, first seen Nov 9, 2026
- Yeachan-Heo/oh-my-claude-sisyphustheir own version, first seen Oct 8, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- Yeachan-Heo/oh-my-claudecode★ 41ksame name
- Yeachan-Heo/oh-my-claude-sisyphus★ 41ksame 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 repository23 other files it gives its agents.
- rules/AGENTS.mdAGENTS.md
- ai-slop-cleanerSkill
- analyzeSkill
- askSkill
- autopilotSkill
- cancelSkill
- code-reviewSkill
- deep-interviewSkill
- deep-researchSkill
- doctorSkill
- hudSkill
- modernizeSkill
- performance-goalSkill
- prSkill
- ralplanSkill
- releaseSkill
- skillSkill
- teamSkill
- threat-modelSkill
- ultragoalSkill
- ultraqaSkill
- visual-ralphSkill
- wikiSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 1 is distinctive, and 6 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
6
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 8
When a design question stalls in prose—two plausible models, neither falsifiable by talk—build a cheap, disposable runnable artifact so the …
- Very specific: line 15
Do not use $loft when the answer is discoverable from repo code (explore), upstream docs (researcher), or a direct Socratic question …
- Very specific: line 29
Rubric First (
arena): before building variants, name the 3–5 concrete evaluation criteria (e.g., scan speed, keyboard density, error … - Very specific: line 36
When N=2..3 competing architectural, algorithmic, state-machine, or UX approaches would collide inside one file and require isolated …
- Very specific: line 43
Disposable by design: no database migrations, no new dependencies, no production wiring, and no commits to main.
- Very specific: line 44
Fast feedback: use ask_question after presenting the loft artifact so the user can select the winning variant or edge-case behavior in one …
- Very specific: line 45
Fold the answer: write the user's chosen option and 1-line rationale into the active .omagy/specs/deep-interview-<slug>.md, …
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Loft — Shape Before Steel ($loft)
When a design question stalls in prose—two plausible models, neither falsifiable by talk—build a cheap, disposable runnable artifact so the user can react to it, then fold their decision back into the spec or plan.
When to use
- A
$deep-interviewor$ralplandecision fork stalls in words ("does this state machine feel right?", "steps, cards, or split-pane?"). - The user invokes
$loftor says "let's prototype both options" / "we'd have to see it first".
Do not use $loft when the answer is discoverable from repo code (explore), upstream docs (researcher), or a direct Socratic question ($deep-interview), or when diagnosing a bug (debugger / $analyze).
Choose the fork
Pick exactly one fork per loft (one loft answers one question):
1. Logic Fork — "Is the state/rule model right?"
For questions about state machines, reducers, scoring formulas, or edge-case transitions:
- Isolate the decision-bearing core as a pure, zero-I/O module (no framework, no network, no persistence).
- Wrap it in a single self-contained clickable HTML harness under
.omagy/artifacts/loft/<slug>/index.html(and optionally copy to the conversation artifact directory and embed inline via<agent-embed src="file:///<artifact_path>/loft-<slug>.html"></agent-embed>when compact<500px, using Antigravity'shttps://www.gstatic.com/antigravity/web/dev/tailwindcss.min.jsand semantic tokens--card,--foreground,--borderper the built-ingenerative_uiskill): one button per scenario/edge case, with the full state transition visible after every click. - Once confirmed, record the chosen state rules in the spec/plan; the pure function can be lifted into production work during
$ultragoal.
2. UI Fork — "What structure holds the data best?"
For questions of layout, information architecture, navigation model, or density:
- Rubric First (
arena): before building variants, name the 3–5 concrete evaluation criteria (e.g., scan speed, keyboard density, error visibility, mobile collapse) at the top of the harness so variants are judged against explicit goals rather than vibe. - Build three variants (
A,B,C) that differ in structure (information hierarchy, navigation flow, density)—never three color/font reskins. - HTML Data-Visualization Hygiene: when a loft prototype renders charts, timelines, or metrics, compute aggregates at the source, bucket dates by the reader's local midnight, hatch partial-window bars so incomplete periods never look like drops, cross-check KPI totals against a second source, use HTML/CSS bars over stretched
preserveAspectRatio="none"SVGs, and scope styles with component-prefixed CSS classes (kpi-label,bar-seg) soA/B/Cnever collide. - Provide an on-page switcher cycling
A/B/Cusing realistic data density in.omagy/artifacts/loft/<slug>/index.html, and surface[oh-my-agy HUD](sidecar://oh-my-agy/hud/)so the user can test the full-heightA/B/Cvariants in theLofttab of the Auxiliary Pane (or open on an isolated git branchloft/<slug>). - Variant Grafting (
arena): once the user picks a winner (or requests a hybrid, e.g., "Variant B's layout with Variant A's filter bar"), graft the winning structural traits intoDESIGN.mdor.omagy/specs/and leave the throwaway harness behind.
3. Multi-Worker Arena Bake-Off ($loft --arena)
When N=2..3 competing architectural, algorithmic, state-machine, or UX approaches would collide inside one file and require isolated runnable prototypes judged against a frozen contract, follow references/arena-bakeoff.md:
- Freeze
.omagy/artifacts/loft/<slug>/contract.mdand shared benchmark/fixture inputs first. - Dispatch
2..3parallelexecutor(ordesigner) subagents withWorkspace: "share"into.omagy/artifacts/loft/<slug>/variant-a/..variant-c/. - Score candidates blindly via an independent
criticsubagent (Lean:LOC, correctness, performance/UX, andcost_if_wrong), recordDecision:/Rejected:in.omagy/artifacts/loft/<slug>/arena-verdict.md, and clean up losing worktrees after preserving evidence.
Discipline
- Disposable by design: no database migrations, no new dependencies, no production wiring, and no commits to
main. Keep artifacts in.omagy/artifacts/loft/<slug>/or on a non-mergedloft/<slug>branch. - Fast feedback: use
ask_questionafter presenting the loft artifact so the user can select the winning variant or edge-case behavior in one click. - Fold the answer: write the user's chosen option and 1-line rationale into the active
.omagy/specs/deep-interview-<slug>.md,.omagy/plans/prd-<slug>.md, orDESIGN.md, then return to the calling workflow.
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.

