m1nga-skill

TaskDock plans and maintains portable Desktop task folders with durable identity, current progress, decisions, and file checks so work can resume after a move.

AI work review pack that collects session files and decisions into one Desktop folder, then verifies the package is complete, linked, and ready for human review.

Marketing prompt builder that turns rough ideas or dictated notes into reusable AI instructions grounded in your brand voice, channel, audience, and source material.

First-time user tester that experiences your product with no prior context and returns a severity-ranked report showing what confuses users and what to fix.

Prompt cleaner for dictated or mixed-language notes. Produce a reusable AI prompt while preserving corrections, examples, scope, and calendar-week boundaries.

Review a skill from the user’s perspective, learn from comparable skills, test improvements, and publish a verified GitHub product and install package.

Product idea stress test that simulates first contact across varied users, exposes the biggest risks, and returns a stronger one-page concept before you build.

LLM judge prompt builder that turns one subjective quality standard and labeled examples into a reliable binary Pass/Fail evaluator prompt.

Iteration cleanup and handoff that preserves current decisions, removes superseded work safely, verifies a cold takeover, and prepares the next milestone.

AI instruction stress test that simulates users and environments to expose trigger misses, skill collisions, stale assumptions, and silent behavior failures before release.

LLM judge validator that compares AI ratings with human labels, measures TPR and TNR, corrects bias, and reports whether production scores can be trusted.

Project lessons log that records costly incidents, distills reusable lessons, and prevents expired decisions from contaminating the next project.

Writing voice matcher that extracts a measurable fingerprint from your own samples and checks AI drafts so they sound like you instead of generic model prose.

Project rebuild advisor that inspects a confused or inherited project and tells you whether to recover, clarify, repair, rebuild, continue, or stop.

Build plan interviewer that asks one focused question at a time and turns a vague idea into a scoped, build-ready plan with explicit non-goals.

Side task assistant that captures bounded background work without interrupting the main conversation, then returns a result or an honest queued status.

Product system map for ideas and codebases: connect user journeys, capabilities, responsibilities, missing paths, and buildable delivery slices at the scope you need.

Decision thinking partner that clarifies ambiguous choices, preserves real alternatives, and recommends a direction only when evidence or values support it.

Product definition interviewer that clarifies who the product serves, what job it does, when and how it is used, why it matters, and what remains unproven.

Retired: replaced by EarBrief, a standalone app that turns work documents into Chinese or English audio. See m1nga/earbrief-app.