Skip to main content
在 Manus 中运行任何 Skill
一键导入
prtamil
GitHub 创作者资料

prtamil

按仓库查看 1 个 GitHub 仓库中的 3 个已收集 skills。

已收集 skills
3
仓库
1
更新
2026-05-13
仓库分布

Skills 分布在哪些仓库

按已收集 skill 数展示主要仓库,并显示它们在该创作者目录中的占比和职业覆盖。

仓库浏览

仓库与代表性 skills

comment-refactor
软件开发工程师

Trigger for the COMMENT_REFACTOR procedure on a phase-2-validated C file whose CODE is already in good shape (pseudocode-shaped from a prior refactor pass) but whose COMMENTS / PROSE need to be rebuilt. Invoke when the user types `COMMENT_REFACTOR`, `UPDATE_LITERATE` (legacy alias), `/comment-refactor <filename>`, or asks to "redo the comments on X". Defaults to the DEFAULT_REFACTOR profile (the competent-learner standard); pass `novice` as an arg (`/comment-refactor novice <file>`) to target NOVICE_REFACTOR instead. Erases every existing narrative comment block, then writes the chosen profile from zero against a code-only inventory. Code bodies are NOT touched.

2026-05-13
default-refactor
软件开发工程师

Trigger for the DEFAULT_REFACTOR doctrine on a validated phase-2 C file — the default choice for competent learners. Invoke when the user types `DEFAULT_REFACTOR`, `/default-refactor <filename>`, or asks to refactor a file for learning where the reader is a competent programmer (knows C, knows general programming, may not know the domain). Concept density + pseudocode-first instead of analogies and worked examples. Heavy doc on data structures + driver functions; light everywhere else. ~30% less prose than `novice-refactor` and better suited to anyone who isn't a true beginner. For absolute beginners who need analogies and worked examples, use `novice-refactor` instead.

2026-05-13
novice-refactor
软件开发工程师

Trigger for the NOVICE_REFACTOR doctrine on a validated phase-2 C file — the textbook treatment for true beginners. Invoke when the user types `NOVICE_REFACTOR`, `COMPLETE_LITERATE` / `LITERATE` (legacy aliases), `/novice-refactor <filename>`, or asks to refactor a file so a first-time reader (no domain background) can learn the algorithm from zero. Reserve for the 5–10 canonical algorithm references in the project. Turns a working file into an embedded textbook via 3 sequential steps — (Step 0) color/HUD/themes, (Step 1) pseudocode-shaped code with extracted helpers, (Step 2) the full prose layer (file header + CONCEPTS + MENTAL MODEL with worked example + GUIDED TUTORIAL with construction-sequence lessons + Tier-1 function labels). Strictly clean-slate — read code only, never existing prose. For competent-learner files (the default case), use `default-refactor` instead.

2026-05-13
已展示 1 / 1 个仓库
已展示全部仓库