원클릭으로 Manus에서 모든 스킬 실행

sigma-claude-md

스타3

포크1

업데이트2026년 7월 10일 14:59

Check or scaffold CLAUDE.md / CLAUDE.local.md against best-practice research (official Anthropic guidance + community consensus + real-world examples). Use when running /claude-md-check or /claude-md-create, when a human asks to "check", "audit", "improve", or "review" a CLAUDE.md file, or when scaffolding a new one and native /init would produce unbounded, undisciplined output. Carries the length/structure/specificity rubric both commands share.

설치

Codex 또는 Claude로 설치 이 Prompt를 복사해 Codex, Claude 또는 다른 어시스턴트에 붙여 넣으면 Skill 페이지를 검토하고 설치를 진행할 수 있습니다.

Manus에서 실행

출처

navidgh66

navidgh66/sigma

GitHub 저장소 열기 Creator 저장소 보기

다운로드

Manus에서 실행

sigma-claude-md

The CLAUDE.md rubric — what a best-practice check or scaffold enforces, distilled from a deep-research pass (2026-07-10) across official Anthropic docs, community practitioner consensus, and 5 real-world repos (cloudflare/workers-sdk, vercel/ai, supabase, humanlayer, langchain).

The one law underneath every check: for each line, would removing it cause Claude to make a mistake? If not, it doesn't belong. Bloat is the dominant documented failure mode — a long CLAUDE.md doesn't fail loudly, Claude silently discounts parts of it (Claude Code injects a system-reminder around CLAUDE.md content saying "you should not respond to this context unless it is highly relevant" — that's the mechanism, not just an anecdote).

Length thresholds

Official target: under 200 lines. Community consensus converges on the same number, with a ~300-line outer ceiling before adherence measurably degrades. >300 = HIGH, 200-300 = MEDIUM.
Nested/per-package CLAUDE.md files are a NATIVE Claude Code mechanism for monorepos, not a workaround — a subdirectory's CLAUDE.md loads lazily, only when Claude reads a file there. Prefer nesting over one giant root file.

Include / exclude (official table, verbatim)

✅ Include	❌ Exclude
Bash commands Claude can't guess	Anything Claude can figure out by reading code
Code style rules that differ from defaults	Standard language conventions Claude already knows
Testing instructions and preferred test runners	Detailed API documentation (link to docs instead)
Repository etiquette (branch naming, PR conventions)	Information that changes frequently
Architectural decisions specific to your project	Long explanations or tutorials
Developer environment quirks (required env vars)	File-by-file descriptions of the codebase
Common gotchas or non-obvious behaviors	Self-evident practices like "write clean code"

Formatting

File:line pointers, never pasted code — a snippet goes stale, a pointer doesn't. Any fenced code block over ~15 lines is a smell.
Imperative voice — "use X" not "we generally prefer X".
WHAT / WHY / HOW structure — tech stack & layout (WHAT), purpose (WHY), workflow & verification steps (HOW). Peer-reviewed evidence (arXiv:2602.11988) found repo-overview/architecture content specifically does NOT improve task success despite being the most common auto-generated content — don't over- invest in WHAT at the expense of HOW.
@path imports are organizational only — they do NOT reduce context (an imported file still loads at launch). A broken import is a real defect (resolves relative to the importing file, not cwd, per official docs).

Register

CLAUDE.md is written FOR an agent, not for a human onboarding to the repo. README voice (narrative history, marketing framing, "in conclusion") is a register mismatch even when the facts are true.

Maintenance

Treat it as a living document: update the moment Claude repeats a mistake twice, prune the moment a rule becomes redundant. Commit to git, review changes like code. CLAUDE.local.md (gitignored) is for personal-only preferences — never duplicate what belongs in the shared file.

Composes with

/claude-md-check — runs deterministic checks (length, pasted code, broken imports, stale test/line counts, placeholders) + this qualitative rubric against CLAUDE.md (required) and CLAUDE.local.md (optional, skipped if absent).
/claude-md-create — scaffolds a WHAT/WHY/HOW starter capped under 200 lines, distinct from native /init (no length/structure discipline there). Refuses to overwrite an existing file.
sigma setup-repo — offers /claude-md-create when neither file exists yet, or surfaces a /claude-md-check finding when one already does (same exists→check / missing→offer-create shape as the learn-artifacts step).

이 저장소의 다른 Skills

같은 저장소

sigma-cost

navidgh66/sigma

Estimate, measure, and optimize token cost for sigma's heavy operations (review's three axes, the profile walk, loop cycles, multi-model research). Use before running a heavy op to size it and pick model tiers, and after to record actual spend and report trends. Triggers: "how expensive", "estimate cost", "which model should this axis use", "what's burning tokens", "sigma cost", or before any multi-axis review / loop / deep research run.

2026-07-043

sigma-docs

navidgh66/sigma

Generate EXTERNAL-facing project documentation from the codebase itself — README, API/reference docs (from code signatures + docstrings), CHANGELOG / release notes (from git log), and an "about this project" presentation. Grounds in ARCHITECTURE.md when present; delegates HTML/theming for presentations to sigma-present. Use when the user wants docs for an AUDIENCE (users, contributors, release readers) — not internal onboarding. Triggers: "write a README", "generate API docs", "draft release notes", "make a project overview deck", "document this for users".

2026-07-043

code-simplifier

navidgh66/sigma

Use when refining already-verified code to fight AI slop — over-abstraction, dead code, duplication, needless cleverness — while preserving behaviour exactly. Powers `sigma loop --simplify`. The simplifier is a DISTINCT agent from the implementer; cleanup is re-verified, never a gate.

2026-06-293

sigma-prune

navidgh66/sigma

Find loaded-but-unused MCP servers and plugins that bloat every Claude context, and reversibly disable the dead weight to save tokens + cost. Use when the context feels heavy, tool lists are huge, or you want to trim what's loaded. Triggers: "too many tools", "context is bloated", "what MCP/plugins am I not using", "save context", "prune unused", "sigma prune", or a periodic context-hygiene pass.

2026-06-253

sigma-scout

navidgh66/sigma

Discover skills relevant to sigma's domains on skillsmp.com and keep the bundle fresh, without ever auto-installing. Use to find trending or new community skills for classic-ml / deep-learning / nlp / rl / data / agents / mlops / llm work, vet them, and pull the keepers in. Triggers: "find new skills", "what skills should sigma add", "scout skillsmp", "keep the bundle current", "any trending skills for <domain>", or a periodic bundle refresh.

2026-06-253

sigma-grilling

navidgh66/sigma

Adversarially grill a sigma design (architecture.md) or specification (spec.md) BEFORE the pipeline advances to code — catch logic flaws, ambiguity, untestable acceptance criteria, missing edge/error paths, scope creep, and ML/data risks while they still cost a sentence to fix instead of a rewrite. Use when running /grill, when a human or the loop wants to pre-flight a spec or blueprint, when about to decompose a spec into tasks, or when the user asks to "review", "pressure-test", "stress-test", "red-team", or "poke holes in" a design/spec. Carries the interrogation rubric; the griller is a separate agent from the author.

2026-06-253

name	sigma-claude-md
description	Check or scaffold CLAUDE.md / CLAUDE.local.md against best-practice research (official Anthropic guidance + community consensus + real-world examples). Use when running /claude-md-check or /claude-md-create, when a human asks to "check", "audit", "improve", or "review" a CLAUDE.md file, or when scaffolding a new one and native /init would produce unbounded, undisciplined output. Carries the length/structure/specificity rubric both commands share.
origin	sigma

sigma-claude-md

Length thresholds

Official target: under 200 lines. Community consensus converges on the same number, with a ~300-line outer ceiling before adherence measurably degrades. >300 = HIGH, 200-300 = MEDIUM.
Nested/per-package CLAUDE.md files are a NATIVE Claude Code mechanism for monorepos, not a workaround — a subdirectory's CLAUDE.md loads lazily, only when Claude reads a file there. Prefer nesting over one giant root file.

Include / exclude (official table, verbatim)

✅ Include	❌ Exclude
Bash commands Claude can't guess	Anything Claude can figure out by reading code
Code style rules that differ from defaults	Standard language conventions Claude already knows
Testing instructions and preferred test runners	Detailed API documentation (link to docs instead)
Repository etiquette (branch naming, PR conventions)	Information that changes frequently
Architectural decisions specific to your project	Long explanations or tutorials
Developer environment quirks (required env vars)	File-by-file descriptions of the codebase
Common gotchas or non-obvious behaviors	Self-evident practices like "write clean code"

Formatting

File:line pointers, never pasted code — a snippet goes stale, a pointer doesn't. Any fenced code block over ~15 lines is a smell.
Imperative voice — "use X" not "we generally prefer X".
WHAT / WHY / HOW structure — tech stack & layout (WHAT), purpose (WHY), workflow & verification steps (HOW). Peer-reviewed evidence (arXiv:2602.11988) found repo-overview/architecture content specifically does NOT improve task success despite being the most common auto-generated content — don't over- invest in WHAT at the expense of HOW.
@path imports are organizational only — they do NOT reduce context (an imported file still loads at launch). A broken import is a real defect (resolves relative to the importing file, not cwd, per official docs).

Register

CLAUDE.md is written FOR an agent, not for a human onboarding to the repo. README voice (narrative history, marketing framing, "in conclusion") is a register mismatch even when the facts are true.

Maintenance

Composes with

/claude-md-check — runs deterministic checks (length, pasted code, broken imports, stale test/line counts, placeholders) + this qualitative rubric against CLAUDE.md (required) and CLAUDE.local.md (optional, skipped if absent).
/claude-md-create — scaffolds a WHAT/WHY/HOW starter capped under 200 lines, distinct from native /init (no length/structure discipline there). Refuses to overwrite an existing file.
sigma setup-repo — offers /claude-md-create when neither file exists yet, or surfaces a /claude-md-check finding when one already does (same exists→check / missing→offer-create shape as the learn-artifacts step).