feature-list-mind
Feature List Mind
A session continuity protocol for long-running agent work. When a project spans multiple context windows, agents lose memory between sessions — leading to over-ambition (attempting everything at once), premature completion (declaring done too early), or repeated work. This skill solves that by establishing a JSON feature list as the single source of truth, a session init sequence that recovers state, and an incremental commit discipline that keeps progress auditable.
The core loop: Resume → Pick Feature → Implement → Verify (steps + test suite) → Commit → Inform User → Update Tracker (after authorization) → Commit Again.
The core structure: Phases → Features → Verification Steps. Phases group features into delivery milestones (Foundation, MVP, Core, Polish). Features are individually completable units of work. Verification steps are the concrete test script for each feature.
Two Modes
The skill operates in two modes, determined by whether features.json exists in the project root.
Init Mode
When features.json does not exist, this is a new project. The goal is to expand the user's project description into a granular, testable feature list and establish the tracking infrastructure.
Read references/init-protocol.md for the full initialization sequence: gathering requirements, expanding into features, creating PROGRESS.md, making the initial git commit, and presenting the feature list for user review.
Resume Mode
More from moghenry/superminds
best-minds-optimizer
Prompt optimizer that rewrites the user's input through the lens of the world's top domain expert before executing. Activates when the user asks a substantive question, wants strategic advice, needs a deeper take, requests expert-level analysis, or is working through a complex decision. Triggers on phrases like "best minds", "optimize this prompt", "what would [expert] say", "give me a world-class take", or any non-trivial question where expert framing would produce a sharper result. Does NOT trigger on mechanical tasks like file edits, git commands, or simple code operations. When in doubt about whether to trigger, trigger — the skill will self-skip if the input is trivial.
104d-mind-analyst
Analyze any problem, phenomenon, or topic from four independent perspectives simultaneously using parallel agents. Use when the user wants multi-dimensional analysis, needs to understand a problem deeply before acting, or says "analyze this", "break this down", "what's really going on with X", "look at this from different angles", or any request for comprehensive analysis of a trend, product idea, social phenomenon, or complex situation. Even if the user doesn't explicitly ask for multi-perspective analysis, trigger this skill when the topic clearly benefits from examining human, product, cultural, and educational dimensions together.
5