ワンクリックでManusで任意のスキルを実行

author-prompt-surface

スター1

フォーク0

更新日2026年7月11日 07:49

Use when creating or editing .claude/agents/*.md, .claude/skills/*/SKILL.md, or .claude/commands/* — apply docs/prompting-standards.md (trigger conditions, model-agnostic bodies, no reasoning-as-text, effort floors).

インストール

Codex または Claude でインストールこの Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。

Manusで実行

ソース

cwijayasundara

cwijayasundara/claude_harness_eng_v5

GitHub リポジトリを開く Creator のリポジトリを見る

ダウンロード

Manusで実行

Author Prompt Surface

Claude-only discipline for the harness monorepo (and any plugin that ships agents/skills). Load before writing or rewriting a prompt surface.

Required reading

Read and follow docs/prompting-standards.md in full. That document is the source of truth; this skill is the invocation trigger + checklist.

Surfaces this skill covers

Path	Role
`.claude/agents/*.md`	Subagent prompts (`model:` frontmatter only for model pins)
`.claude/skills/*/SKILL.md`	Skill prompts and stage workflows
`.claude/commands/*`	Slash-command entrypoints

Checklist (every new or edited prompt)

Copy from docs/prompting-standards.md — fail the edit if any box is open:

No anti-laziness CRITICAL / MUST / If in doubt left in (true invariants only).
Every tool/subagent has a use this when … trigger condition.
Finding/review steps say report everything with severity; gates filter downstream.
Long-running steps audit progress claims against tool results.
No "show/echo/transcribe your reasoning as text".
Distinct blocks in XML tags; examples where behavior is subtle.
Effort expectation noted for agentic/coding work (high / xhigh).
No model named in the prompt body; no directional nudge that assumes one model's default (criterion, not nudge).

Model pins

Pin models only in agent frontmatter (model:) or the session model.
Never name a model generation in the prompt body.
One prompt body must run unchanged if the pin changes.

When not to use this skill

Product application code under src/ / generated apps.
Disposable artifacts (mockups, ARB narratives, research) — use the lite lanes.
Editing deterministic hooks/scripts — use code-gen + tests, not prompting standards.

After editing

If the change is a new control (gate, sensor, reviewer), also follow HARNESS.md / harness-manifest.json registration so the control is not orphaned.

このリポジトリの他の Skills

同じリポジトリ

auto

cwijayasundara/claude_harness_eng_v5

Autonomous build loop with Karpathy ratcheting, GAN evaluator, and session chaining. Iterates story groups until all features pass or stopping criteria met.

2026-07-121

change

cwijayasundara/claude_harness_eng_v5

Change the behavior of existing code — story-driven by default, or --issue N for a GitHub bug fix. Test-first, full verification, code review.

2026-07-121

code-gen

cwijayasundara/claude_harness_eng_v5

Code generation quality principles — TDD, typing, error handling, logging, API integration, LLM integration.

2026-07-121

feature

cwijayasundara/claude_harness_eng_v5

Brownfield change route — take an existing-code feature request from intent to a reviewed PR, scaling from a single /change to an epic via /spec→/design→/auto. Linear-tracked, backed by a committed DeepWiki.

2026-07-121

fix-from-diagnostics

cwijayasundara/claude_harness_eng_v5

[Internal pipeline stage — run by /auto self-heal and /implement validation when lint/type error volume is high; invoke directly only as a power user.] Turn compiler/linter output into a sharded work queue; fix per package with optional adversarial review — no full monorepo suite mid-shard.

2026-07-121

implement

cwijayasundara/claude_harness_eng_v5

[Internal pipeline stage — run by /auto (follow with /gate); invoke directly only as a power user.] Generate production code and tests for a story group using agent teams for parallel execution.

2026-07-121

name	author-prompt-surface
description	Use when creating or editing .claude/agents/.md, .claude/skills//SKILL.md, or .claude/commands/* — apply docs/prompting-standards.md (trigger conditions, model-agnostic bodies, no reasoning-as-text, effort floors).
user-invocable	false

Author Prompt Surface

Claude-only discipline for the harness monorepo (and any plugin that ships agents/skills). Load before writing or rewriting a prompt surface.

Required reading

Read and follow docs/prompting-standards.md in full. That document is the source of truth; this skill is the invocation trigger + checklist.

Surfaces this skill covers

Path	Role
`.claude/agents/*.md`	Subagent prompts (`model:` frontmatter only for model pins)
`.claude/skills/*/SKILL.md`	Skill prompts and stage workflows
`.claude/commands/*`	Slash-command entrypoints

Checklist (every new or edited prompt)

Copy from docs/prompting-standards.md — fail the edit if any box is open:

No anti-laziness CRITICAL / MUST / If in doubt left in (true invariants only).
Every tool/subagent has a use this when … trigger condition.
Finding/review steps say report everything with severity; gates filter downstream.
Long-running steps audit progress claims against tool results.
No "show/echo/transcribe your reasoning as text".
Distinct blocks in XML tags; examples where behavior is subtle.
Effort expectation noted for agentic/coding work (high / xhigh).
No model named in the prompt body; no directional nudge that assumes one model's default (criterion, not nudge).

Model pins

Pin models only in agent frontmatter (model:) or the session model.
Never name a model generation in the prompt body.
One prompt body must run unchanged if the pin changes.

When not to use this skill

Product application code under src/ / generated apps.
Disposable artifacts (mockups, ARB narratives, research) — use the lite lanes.
Editing deterministic hooks/scripts — use code-gen + tests, not prompting standards.

After editing

If the change is a new control (gate, sensor, reviewer), also follow HARNESS.md / harness-manifest.json registration so the control is not orphaned.