HelloWorldU
UserBattle-tested agent skills on judgment before action, distilled from real OSS work
Categories
Indexed Skills (3)
progressive-abstraction
Three-layer issue triage methodology for unfamiliar codebases — never dive into implementation details before judging the fix at the architecture layer. Layer 1 identifies the owning layer and classifies the problem (correctness vs completeness vs product decision), Layer 2 traces the real data flow with verified file:line evidence and surfaces boundary semantics, Layer 3 proposes bounded solutions that declare what they deliberately do not do. Use when analyzing or triaging an issue/bug in an unfamiliar repository, deciding where a fix belongs before writing code, preparing an OSS contribution (issue comment / PR) that must survive maintainer review, or evaluating third-party and AI-generated review opinions against code evidence. For vague greenfield ideas use idea-to-spec first; for concrete modification tasks in a familiar repo use task-intake instead.
task-intake
Mandatory alignment gate before modifying code, files, or configuration — restate the requirement, surface assumptions, enumerate scenarios with expected outcomes, then STOP and wait for explicit user approval before writing any implementation code. Use at the start of any task that involves changing code, files, or configuration in a repository, especially multi-file or multi-step changes. Skip the gate only for pure Q&A, pure explanation, or tiny changes where the user explicitly said "just do it" — the implementation invariants still always apply. For vague greenfield ideas that are not yet concrete tasks, use idea-to-spec first; for triaging issues in unfamiliar codebases, use progressive-abstraction.
idea-to-spec
Turn a vague idea into a concrete, actionable spec through structured dialogue — disambiguating requirements, surfacing decision points, and probing technical feasibility before any code is written. Use when the user brings an open-ended or half-formed idea ("I want to build...", "what if we...", "help me create...") and the shape of the thing is not yet clear. Converges into an approved spec, after which concrete modification tasks should go through task-intake, and unfamiliar-codebase investigation through progressive-abstraction. Do not write code inside this skill — producing the aligned spec and getting it approved is the entire deliverable.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.