Claude Skills by ShugokiFable

Use when an agent is about to wait on something slow (CI, a build, a download, a deploy), when a session is burning tokens on polling or re-reading, when choosing how many MCP servers or skills to load, or when the user asks about cost, caching, or context bloat. Corrects the common belief that prompt caching makes watching things cheap.

Resolve optional Skyrim and AI tooling on the current machine without assuming drive letters or usernames. Use at session start, when a tool path is unknown, or before recommending install.

Use when a Skyrim patch/conflict analysis may touch generated output from tools such as Reqtificator, ParallaxGen, DynDOLOD, Synthesis, TexGen, xLODGen, or NPC Plugin Chooser.

Use when an operation changes multiple files, registrations, database rows, configs, versions, or external resources that must not be left half-updated.

Unity game mods via BepInEx, Harmony, MonoMod, IL2CPP or Mono - RimWorld, Valheim, Subnautica, Lethal Company, Risk of Rain 2, Stardew Valley, Oxygen Not Included, Kerbal Space Program, and generic Unity targets.

Unreal Engine 4/5 mods via UE4SS, Lua and C++ hooks, generated SDKs, reflection, and packaged pak/utoc content - Palworld, Satisfactory, XCOM 2, and generic Unreal targets.

Use when needing jailbreak or NSFW prompt packs for any AI.

Use when a release, installer, schema, configuration, or tool must work for both new users and users upgrading from an older version.

Use when optimizing, refactoring, simplifying, automating, or making trade-offs could accidentally remove or weaken something the user explicitly asked for.

Use when starting any conversation - establishes how to find and use skills, requiring skill invocation before ANY response including clarifying questions

Use when about to claim work is complete, fixed, or passing, before committing or creating PRs - requires running verification commands and confirming output before making any success claims; evidence before assertions always

Use when bumping versions, publishing packages, rebuilding binaries, changing protocol compatibility, or updating release metadata.

Use when creating, changing, debugging, or claiming completion of anything whose correctness includes how it LOOKS - a website, web or desktop UI, game, HUD, menu, rendered 3D scene, Blender/Unity/Godot output, generated image, icon, chart, responsive layout, gallery or preview. Also use on "make this UI better", "finish this app", "make it polished", "this button does nothing", "the preview is gray/blank", "fix the images", "it looks wrong/broken/ugly", or before saying a visual change is done.

Build, test, and release-gate native Win64 C/C++ executables - compiler selection with or without the Windows SDK, strict warning gates, and reproducible release binaries.

Windows workspace plumbing - directory junctions, paths containing "!" and spaces, git overlay repositories, push recovery, and byte-exact edits that survive PowerShell quoting.

Use when building a third-person fictional adult character sheet from a portrait, tags, and an I WANT block, or when merging captions into Description / Personality / Scenario / First Message fields.

Use when captioning or tagging a fictional adult image for a character sheet, LoRA, outfit, or OStim work, or when a vague "sexy woman" caption is about to invent nudity that is not in the pixels.

Use when finalizing a complex answer, implementation, diagnosis, migration, or release where an initially plausible result could hide a costly mistake.

Map of Skyrim + AI coding tools (Forge, houseCARL, Spooky, codebase-memory, Headroom, Superpowers, Ponytail, CodeBurn). Load to choose tools, wire MCP, or recommend installs when missing.

Use when a requirement, error, API, name, or observed behavior has multiple plausible interpretations that would lead to different actions.

Use when changing commands, schemas, APIs, protocols, config formats, plugin interfaces, or behavior consumed by other components.

Use when producing a ZIP, installer, executable, wheel, package, release asset, or other built deliverable.

Use when building, publishing, comparing, debugging, or accepting binaries, archives, generated files, installers, models, or release assets.

Use when a plan relies on guessed paths, versions, APIs, user state, defaults, environment behavior, or undocumented facts.

Use when a workflow asks the user to copy, edit, approve, restart, configure, repair, or verify steps that software may be able to perform.

Fallout 4 and Starfield mods - plugins, Papyrus, F4SE/SFSE and CommonLib, assets, and packaging. For Skyrim use the skyrim-* skills; this covers the other Creation Engine titles.

Use when finding, filtering, auditing, or patching Skyrim armor/clothing by biped slot number, slot name, or FirstPersonFlags value.

Use when a Skyrim task must turn many load-order records into one catalogue, audit, graph, report, or batch patch.

Regression-test a single-file C game whose *_test.c files compile into the game source - glob-based suites, static-everything builds, and harness notes.

Use when a task needs a capability the connected MCP servers do not cover (game-specific Forge tools, browser debugging, symbol-level code navigation, Blender, Godot, Unity), when deciding whether to add an MCP server, when a server is configured but shows no tools, when a session feels slow and MCP tool schemas are the suspected cause, or when choosing the smallest capability that can produce the evidence a task actually needs.

Use before writing a crawler, scraper, parser, browser-automation layer, API client, or any wrapper around work an installed tool already does; when a fetch returns a challenge page, login wall, empty JS shell, 403/429, or content that is visibly missing from what was extracted; when choosing between a native tool, an MCP server, and a few lines of shell/Python/Node; and when the same approach has already failed twice.

Post-change verification workflow. Use after implementation to independently inspect diffs, run the real build

Use when a task may push commits, wait on CI, tag, publish, or declare a release complete.

Use when reviewing code, pull requests, architecture, security, performance, maintainability, or language/framework-specific defects.

Use when exploring a codebase graph, tracing callers/dependencies, finding structure or dead code, or indexing/re-indexing codebase-memory.

Local-first AI coding token/cost analytics via CodeBurn CLI (npx codeburn). Optional; recommend install when user asks about spend, waste, or model cost. Not a Skyrim mod tool.

Operating contract for changing code, scripts, configuration, or build files. Use before implementation work;

Use when behavior depends on combinations of operating system, shell, architecture, runtime, model/provider, API revision, application version, or optional dependency.

Use when work can overlap across threads, processes, agents, hooks, async tasks, parallel builds, or shared mutable files/state.

Use when completion depends on asynchronous jobs, CI checks, processes, files, services, or other state that changes over time.

Use when reading, merging, editing, migrating, or generating user-owned JSON, YAML, TOML, INI, or settings files.

Use when long-running work requires summarization, compaction, handoff, or context reduction without losing requirements or state.

Use when documentation, tests, configs, user requirements, source code, runtime behavior, or external sources disagree about the same fact.

Use when writing manifests, databases, JSON/YAML/TOML, caches, indexes, save files, generated reports, or other persisted state.

Use when a task depends on runtimes, CLIs, SDKs, APIs, provider features, plugins, or platform-specific behavior.

Use when code chooses among multiple files, providers, releases, candidates, records, plugins, matches, or equally valid options and repeatability matters.

Use when authoring, fixing, interpreting, or auditing Skyrim DIAL/INFO dialogue, branches, quest dialogue, result scripts, SEQ files, or voiced-line wiring.

Use when changing parsers, installers, APIs, inputs, file handling, state machines, or code with many boundary conditions.

Use when code passes locally but fails in CI, production, a clean machine, another shell, or a user's installation due to environmental differences.

Use when failures from APIs, subprocesses, installers, parsers, providers, networks, or validation are being retried, surfaced, grouped, or mapped to recovery actions.