一键在 Manus 中运行任何 Skill

gf-change-audit

星标23

分支4

更新时间2026年6月3日 12:22

Split large GF Framework changes into focused review tracks before commit, release, or risky integration. Use after broad refactors, API changes, layer moves, extension restructuring, reference-project migration work, or when the user asks for multi-agent audit.

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

C76GN

C76GN/gf-framework

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

GF Change Audit

Use this skill to review large GF changes without mixing concerns. When sub-agents are available, run independent read-only tracks in parallel; otherwise run the same tracks sequentially.

Prepare

Pin the review scope:
- For committed work, ask or infer a base and use git diff <base>...HEAD.
- For dirty work, inspect git status --short --branch, git diff --name-status, and staged diff if present.
List touched files and classify them with python tools\gf_maintenance.py workspace-status --json.
Identify the user's task intent, acceptance criteria, and explicit out-of-scope boundaries. If no spec exists, audit only against repository rules and stated conversation intent.

Parallel Tracks

Use the prompts and output format in review-axes.md. Keep each track bounded and independent:

Standards: AI_MAINTENANCE.md, CODING_STYLE.md, API_SURFACE.md, layer boundaries, generated-doc rules, release rules.
API and compatibility: public API surface, @since, schema docs, SemVer, deprecated/removed behavior, generated reference drift.
Verification: tests, docs checks, examples checks, release checks, skipped commands, and whether failures were expected.
Reference boundary: external gf-reference-project, sync behavior, smoke paths, framework feedback, and project-business leakage.
Architecture: module depth, locality, dependency direction, extension atomicity, and whether a new seam earns its interface cost.

Aggregate

Report blocking findings first, then warnings, then clean tracks. Cite local file paths and commands for concrete review findings. Do not merge away disagreement between tracks; preserve it as an explicit open question.

When a track finds a deterministic rule that is not automated yet, recommend whether it belongs in tests/gf_core/maintenance, tools/gf_maintenance.py, or AI_MAINTENANCE.md.

同仓库更多 Skills

同仓库

gf-framework-maintenance

C76GN/gf-framework

Maintain GF Framework source, docs, tests, API surface, generated references, and maintenance tooling. Use when Codex modifies addons/gf, tests/gf_core, docs/zh, API docs, tools, AI maintenance policy, or needs to choose GF-specific validation commands.

2026-07-2323

gf-project-development

C76GN/gf-framework

Build, refactor, debug, and review Godot projects that use GF Framework. Use for project architecture, GF API selection, lifecycle and ownership decisions, package or extension selection, project-owned platform adapters, validation, and evidence-based feedback to the GF repository.

2026-07-2323

gf-release-flow

C76GN/gf-framework

Prepare and verify GF Framework version releases, Asset Store packages, commits, and SemVer tags. Use when the user asks to determine a version, enter release flow, update release metadata, build release packages, commit, tag, push, or check a GF release.

2026-07-1823

gf-reference-boundary

C76GN/gf-framework

Maintain the external GF reference project boundary and examples smoke workflow. Use when working with gf-reference-project, examples suite, sync_reference_project.py, smoke scenes, reference project feedback, or avoiding project-business leakage into addons/gf.

2026-06-0323

name	gf-change-audit
description	Split large GF Framework changes into focused review tracks before commit, release, or risky integration. Use after broad refactors, API changes, layer moves, extension restructuring, reference-project migration work, or when the user asks for multi-agent audit.

GF Change Audit

Use this skill to review large GF changes without mixing concerns. When sub-agents are available, run independent read-only tracks in parallel; otherwise run the same tracks sequentially.

Prepare

Pin the review scope:
- For committed work, ask or infer a base and use git diff <base>...HEAD.
- For dirty work, inspect git status --short --branch, git diff --name-status, and staged diff if present.
List touched files and classify them with python tools\gf_maintenance.py workspace-status --json.
Identify the user's task intent, acceptance criteria, and explicit out-of-scope boundaries. If no spec exists, audit only against repository rules and stated conversation intent.

Parallel Tracks

Use the prompts and output format in review-axes.md. Keep each track bounded and independent:

Standards: AI_MAINTENANCE.md, CODING_STYLE.md, API_SURFACE.md, layer boundaries, generated-doc rules, release rules.
API and compatibility: public API surface, @since, schema docs, SemVer, deprecated/removed behavior, generated reference drift.
Verification: tests, docs checks, examples checks, release checks, skipped commands, and whether failures were expected.
Reference boundary: external gf-reference-project, sync behavior, smoke paths, framework feedback, and project-business leakage.
Architecture: module depth, locality, dependency direction, extension atomicity, and whether a new seam earns its interface cost.

Aggregate

When a track finds a deterministic rule that is not automated yet, recommend whether it belongs in tests/gf_core/maintenance, tools/gf_maintenance.py, or AI_MAINTENANCE.md.