lunarmoon26/lunarmoon26-agent-skills

Use for repository-level documentation-first engineering when the user explicitly requests document-driven or spec-first work, a PRD or feature contract, canonical documentation ownership, docs-to-code drift repair, synchronized docs/code/tests, or requirement-to-evidence closure. Also use for consequential externally visible or persisted behavior changes that must update an existing contract. Do not use for ordinary coding with no contract impact, one-off prose or file-format editing, architecture briefs or ADRs, or comparative source research.

Use for repository architecture documentation and significant design decisions: architecture briefs, arc42-style context or views, system boundaries, decomposition, critical runtime or deployment flows, quality scenarios, risks, architecture reviews, ADRs, decision records, option tradeoffs, and superseding an accepted decision. Use when the user asks to document or review architecture without a heavyweight research program. Do not use for routine feature documentation or exhaustive comparative codebase research.

Use when the user wants to resize, crop, convert, optimize, recolor, mask, flood fill, use a Magic Wand-style contiguous color selection, make image backgrounds transparent, create transparent icon assets, split a sprite sheet, calculate icon crop offsets, or generate CSS sprite positions for PNG, JPEG, WebP, AVIF, TIFF, BMP, GIF, or ICO files. Use this skill for image/photo/texture/icon asset editing, including when the user does not name a file format.

Use for evidence-led engineering when a task requires drilling into local or external codebases, comparing implementations, frameworks, libraries, or standards, evaluating strengths and weaknesses, tracing how a system really works, or researching before a consequential implementation choice. Use when the user says investigate, compare approaches, survey repositories, learn from existing projects, find prior art, or validate a design against source. Do not use for routine code navigation, a simple documentation lookup, or architecture/spec writing that needs no source research.

Use when writing a Chinese-language article or note that explains a new architecture, framework, runtime, library, protocol, or system design by deriving it from zero instead of listing APIs. Triggers include requests to explain a system step by step in Chinese, write a "from scratch" breakdown, distill how a technology works for a general CS audience, or adapt an existing technical explanation into this derivation style. Do not use for API reference writing, release notes, translation of existing docs, or content aimed at literal children.