Claude Skills by NeverSight

Use this skill to review, tighten, draft, and rewrite prose so it does not read like generic LLM output, AI writing, bland marketing, or generic launch copy. Apply it when the user asks to make writing less generic/AI-sounding, tighten a talk intro, sharpen presentation/slide copy, or improve articles, wiki pages, README text, emails, posts, scripts, product/DevRel copy, launch copy, and other important writing with AI tropes, inflated significance language, generic cadence, weak flow, or mar...

Comprehensive audit toolkit with 14 audit types. Includes a pre-push branch audit (8-category checklist with Clean/Minor/Blocking verdicts) plus 13 deep-dive audits run via sub-agents: code quality, documentation brittleness, docs-code sync, language best practices, concurrency, resource management, test quality, feature completeness, performance, bug patterns, design philosophy compliance, security vulnerabilities, and UI design (CRAP principles). ALWAYS use this skill when the user wants to...

Create and improve README documents for GitHub projects. Use when the user wants to write a new README, improve an existing one, audit README quality, or asks about documentation best practices for their repository.

Code quality verification gates wired into the agent lifecycle. Use this skill whenever writing, modifying, reviewing, or debugging code — including new features, bug fixes, refactors, troubleshooting, CI/CD setup, or project bootstrapping. Also use when the user mentions "quality", "testing strategy", "CI pipeline", "guardrails", "debugging", or asks how to improve code reliability. If you're writing code or trying to understand why code isn't working, this skill applies.

Initialize a new project by creating the .project/ directory structure. Run this first before any other skill. Creates the foundation for all subsequent work.

Remote Linux management for Alma via SSH/SFTP with persistent stateful SSH sessions, thread-session binding, NL-to-command bridge, xterm.js websocket terminal bridge, dangerous-command approval flow, strict command policy option, host-key verification modes, RBAC allowlist, and SIEM-friendly redacted audit fields.

Strict dosage for the adderall meta-skill pack — adherence 0.95, flexibility 0.05. Target skill is executed as a specification; only trivial phrasing freedom remains.

Maximal-literal dosage for the adderall meta-skill pack — adherence 1.00, flexibility 0.00. The target skill is executed verbatim; the agent contributes no interpretation.

Run a bounded 3-cycle interactive review-and-fix workflow against a user-chosen git review target (working-tree diff, current branch vs. auto-detected base, or an explicit commit/tag/branch ref) using the codex plugin. Each cycle invokes codex `review` or `adversarial-review --json`; Claude verifies each finding against a six-item validity checklist, calls `review-scope-guard` for Definition-of-Done triage, and the user picks which findings to fix before the next cycle. Covers both code diffs...

Use when applying user-selected review findings after scope triage, especially inside codex-review-cycle or after a standalone Codex adversarial review, and a line-scoped fix may create follow-on valid findings. Do not use for ordinary edits, plan drafting, single-shot lint, no-review contexts, or generic refactor checklists.

Triage code/plan review findings against an explicit Definition of Done so must-fix bugs are separated from scope creep, out-of-scope semantic implementations, and noise. Collects the six-item Definition of Done interactively on first invocation, classifies every finding into one of four categories (`must-fix`, `minimal-hygiene`, `reject-out-of-scope`, `reject-noise`), maintains a rejected-findings ledger so repeated complaints are not re-litigated across cycles, and evaluates five stop signa...

Use when the user explicitly asks for vibe brainstorming, creative implementation ideas, implicit expected behavior, or convention checks, and when an implementation task is creative or convention-dependent. Do not use for obvious mechanical edits.

Use when the user wants to understand, locate, trace, or assess existing code without changing it — questions like "how does X work", "where is Y implemented", "what would changing Z affect", architecture or data-flow mapping, dependency tracing, convention discovery, or pre-planning and pre-debugging evidence gathering. Do not use when a defect needs repair, a plan or spec artifact is requested, or an edit, commit, or diff review is the deliverable.

Use when the user explicitly invokes vibe-coding through a host-specific skill command, host-provided invocation signal, or direct instruction such as "use `vibe-coding`" for a coding workflow.

Use when writing, revising, reviewing, or critiquing git commit messages, especially commit bodies, Conventional Commits, release commits, dependency updates, monorepo/package changes, i18n/localization commits, performance work, CI/build automation, security/data-loss fixes, agent-authored history, LLM-readable project history, or requests for vibe/AI-friendly commit context.

Use when the user asks to commit, stage, or "save" agent-assisted coding changes — including vague requests like "commit this", "commit please", "コミットして", or "/commit" — and the real work is deciding which files belong in the commit, excluding unwanted or generated files, splitting unrelated changes, or fixing a commit's file set, multi-line message transport, history (amend/rebase), or authorship trailers. This skill owns commit execution, message transport, and git safety.

Use when debugging or repairing existing features from rough agent-assisted coding reports, regressions, failed prior fixes, repeated "still broken" feedback, source-only debugging stalls, unobserved runtime state, tool or automation failures, environment-specific failures, runtime artifact mismatches, security boundary surprises, or fixes that feel wrong.

Use when the user asks to align, confirm, or correct the agent's understanding before action; when prior misinterpretation, risky ambiguity, release/version/commit intent, destructive effects, or goal disagreement could cause rework or damage.

Use when coordinating subagents for coding, research, repair, or review work where delegated workers may drift, stall, crash, duplicate, or edit a shared workspace and the coordinator must preserve scope, verification, and user-consent boundaries.

Use when the user asks to execute, implement, continue, or apply an existing vibe-planning output, implementation plan, specification, acceptance criteria, or task plan. Do not use for plan creation or coding requests with no concrete plan to bind.

Use when the user asks to review, confirm, walk through, or pre-check a saved Markdown implementation plan before implementation; interactively reviews plan items one at a time and manages localized item-level decisions.

Use when the user wants planning before coding: plan mode, create a plan, implementation plan, specification, acceptance criteria, test plan, vibe coding, requirements clarification, what to build first, or equivalent planning requests in any language. Also use for rough, ambiguous, feasibility-sensitive, non-technical, or not-yet-implementable coding requests.

Use when a user wants to draft, revise, save, approve, or explicitly explore requirements for a rough, ambiguous, contradictory, creative, non-technical, or underspecified coding goal before implementation planning or coding, including explicit chat-only/no-file requirements exploration.

Use when the user asks for agent-assisted code review of a git diff, working tree, branch, base ref, git-backed plan or document change, or review/fix loop where scope triage, delegated reviewers, specification gaps, or cascade-safe fixes may matter.

Use when the primary task is writing, revising, reviewing, or critiquing agent-assisted coding development text, source-code comments or docstrings, README/docs, CHANGELOG/release notes, PR descriptions, UI copy, chat replies, progress updates, final summaries, or git commit messages, especially when text must be LLM-readable, meaning-preserving, format-bound, language-aware, or evidence-bound. Treat incidental wording inside another active workflow as auxiliary.

Core macOS automation skill using MacPilot CLI. Enables Claude Code to control apps, type text, click elements, run shell commands, and automate workflows on macOS via the `macpilot` command.

Handle macOS file dialogs (Open, Save, Print) with MacPilot. Navigate folders, select files, set filenames, and dismiss dialogs programmatically in any application.

Capture screenshots and extract text via OCR using MacPilot. Take full-screen, region, or window screenshots, and recognize text in images or screen areas with multi-language support.

Inspect and interact with macOS UI elements using MacPilot accessibility APIs. Find buttons, text fields, labels, and other elements by role, label, or position, then click, read, or modify them.

Manage macOS windows with MacPilot. List, move, resize, snap, minimize, fullscreen, and arrange application windows. Supports multi-display and Spaces.

Use this skill whenever building Flutter widgets, screens, or modifying UI in the Dukaan AI app. Contains the exact color tokens, spacing values, text styles, widget patterns, and UX rules for this project. Ensures all UI follows the Dukaan AI design system and targets low-end Android devices.

Use this skill whenever implementing payment flows, credit deduction, subscription management, or Razorpay integration in Dukaan AI. Contains the exact plans, pricing, credit rules, Razorpay flow, and Worker API contract for payments.

Create a lightweight project spec by identifying the decisions that must be made before implementation, without resolving them. Use as the first step in the spec-driven workflow to produce the initial spec for later exploration and fortification.

Explore a single unresolved decision in an existing project spec by surfacing options, tradeoffs, assumptions, and unknowns without forcing a premature choice. Use as the second step in the spec-driven workflow to deepen and refine the spec one decision at a time.

Examine a project spec after its decisions have been explored to find contradictions, gaps, accidental complexity, and other issues. Use as the final step in the spec-driven workflow to strengthen the spec before implementation begins.

Guide users to build and test their MVP using the BMAD Method with PMF context files. Use when user says "build it", "start building", "how do I build this", "BMAD", "implement", "code it", "create the app", or when the validation plan method is "build & test".

Build the ICP section of your PMF context layer. Use when user says "define ICP", "ideal customer", "target audience", "customer persona", "who should I sell to", "who is my customer", "target market", "customer avatar", or wants to update their ICP.

Create an outreach plan and Mom Test interview questions for conversation-based validation. Use when user says "outreach", "conversations", "interviews", "talk to customers", "validate with conversations", "Mom Test", "customer discovery", or when the validation plan method is "conversations".

Guide users through building their complete PMF context layer. Triggered by /plan-pmf command or when user wants to build full context.

Build your PMF context layer - reference files that capture the "WHY" behind your product. Use when user mentions "PMF", "product market fit", "define my PMF context", "PMF context", "ICP", "value prop", "aha moments", or asks about understanding customers or market positioning.

Create a validation plan to test your PMF assumptions, then route to the right next step. Use when user says "validate", "test my assumptions", "validation plan", "how do I know if this works", or after completing the MVP PRD.

Build the Value Proposition section of your PMF context layer using Callout + Magnet. Use when user says "value proposition", "value prop", "messaging", "positioning", "callout and magnet", "brand message", "why would customers buy", "unique value", or wants to update their value proposition.

Your skill search engine. When you encounter a task you don't know how to do — processing PDFs, deploying to cloud, editing documents, calling unfamiliar APIs, or any specialized workflow — search this marketplace to find a skill that teaches you how. Search by keyword or category, read the skill's instructions, download and use it. Use this whenever you need to learn a new capability, find a tool integration, or discover how to accomplish an unfamiliar task.

Use when the user wants to resume interrupted work, reconstruct context, find buried commitments, explain why a file or tab is still open, prepare a handoff, safely clean Downloads, list or inspect threads, show activity history, search past work sessions, split or merge threads, export/import threads for sharing, or search raw events. Requires a running Threadline MCP server.

Mandatory discipline requiring grep/search before any file modification. Prevents breaking existing patterns by ensuring all usages, imports, and dependencies are understood before edits are applied.

Essential security checklist and patterns for web applications. Use when reviewing code for security issues, implementing authentication, or hardening an application. Covers OWASP top 10, input validation, and secure coding practices.

Use AG Grid MCP-style docs search, version detection, examples, definitions, and upgrade guidance through a lightweight local CLI wrapper. Use when working on AG Grid integration or migration tasks.

Use when auditing Claude skills and commands for quality. Supports Quick Scan (changed skills only) and Full Stocktake modes with sequential subagent batch evaluation. TRIGGER when: "/skill-stocktake", "audit my skills", "skill quality check", "review all skills", "skill inventory", "which skills need work"

Keep a PR moving through review, conflicts, and CI until merge-ready.

Diagnose reproducible bugs using temporary runtime probes and captured evidence.