inprealpha/inprealpha-agent-skills
Personal workflows for engineering, review, investigation, verification, and craft.
Design caller usage, interfaces, data types, and module ownership before implementing a change.
Assess what a change could break beyond its diff and test the assumptions its safety depends on.
Simplify writing, documentation, code, architecture, interfaces, and creative work so people can quickly understand and use it. Use for “AI slop,” confusing or overcomplicated output, project reviews, and making the current result ready to share or ship. Not an AI detector.
Create executable project verification instructions with a feature map and retained evidence.
Improve one measurable outcome through isolated hypotheses, repeated measurements, and regression checks.
Explain a subsystem through traced runtime flow, data ownership, and concrete code references.
Adversarially review a change and deliver a prioritized verdict grounded in reachable failures.
Audit a project verification skill against current source and observed application behavior.
Coordinate substantial work with ownership suited to the task, parallel execution, and fresh perspectives. Use for broad ambiguous goals, separable workstreams, or requested team workflows.
Extract durable lessons from a work session and route them to focused skill or tooling improvements.
Finish work through repeated fresh-context reviews and fixes. Use when the user wants an autonomous review loop after a brief alignment on the outcome and review perspectives.
Simplify code when asked to review over-engineering or apply cleanup, including redundant tests and avoidable dependencies. Preserve behavior; keep unrelated feature work and general correctness reviews outside this skill's scope.
Compare skill variants on realistic tasks with controlled inputs and evidence-based scoring.
Investigate the history and rationale behind code, with cited facts and explicit uncertainty.