csh/playbook
1
1
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

♻️ refactor(agent): streamline skills and rulesets

Browse Source
This commit is contained in:
csh
2026-08-11 14:56:21 +08:00
parent 9010829c6d
commit 34a3a5dec1
13 changed files with 333 additions and 121 deletions
Download Patch File Download Diff File Expand all files Collapse all files
.gitea/ci/thirdparty-skill-overlays/codebase-recon.patch
+85
View File
@@ -0,0 +1,85 @@
diff --git a/skills/thirdparty/codebase-recon/SKILL.md b/skills/thirdparty/codebase-recon/SKILL.md
--- a/skills/thirdparty/codebase-recon/SKILL.md
+++ b/skills/thirdparty/codebase-recon/SKILL.md
@@ -252 +252 @@ Core methodology:
-- [confidence.md](../pathfinding/references/confidence.md) — confidence calibration (shared with pathfinding)
+- [confidence-calibration.md](references/confidence-calibration.md) — evidence-based confidence calibration
@@ -263 +263,0 @@ Related skills:
-- `outfitter:pathfinding` — clarifying requirements before analysis
diff --git a/skills/thirdparty/codebase-recon/references/confidence-calibration.md b/skills/thirdparty/codebase-recon/references/confidence-calibration.md
new file mode 100644
--- /dev/null
+++ b/skills/thirdparty/codebase-recon/references/confidence-calibration.md
@@ -0,0 +1,66 @@
+# Confidence Calibration
+
+Confidence measures how strongly the available evidence supports a codebase
+finding. It does not measure requirement clarity or implementation readiness.
+
+## Calibration Dimensions
+
+Consider these dimensions together; do not average them mechanically:
+
+- **Directness**: runtime observation and source code outweigh inference.
+- **Coverage**: evidence spans the relevant execution paths and modules.
+- **Corroboration**: code, tests, documentation, and history agree.
+- **Counter-evidence**: plausible alternatives and contradictory cases were checked.
+- **Remaining gaps**: unknown behavior is bounded and stated explicitly.
+
+## Typical Levels
+
+| Level | Evidence state |
+| --- | --- |
+| 0 | No repository evidence collected yet. |
+| 1 | Broad structure or search results only. |
+| 2 | Relevant code located, but behavior or scope is not verified. |
+| 3 | Main path traced with some corroboration; meaningful gaps remain. |
+| 4 | Multiple sources support the finding and counter-examples were checked. |
+| 5 | Relevant scope is covered, alternatives are ruled out, and conclusions are directly supported. |
+
+A clear user request can narrow the investigation but cannot raise confidence in
+a finding by itself. Documentation alone normally remains at level 1-2. A
+single code path without tests, history, or runtime confirmation normally
+remains at level 2-3.
+
+## Examples
+
+### Focused Question, Strong Evidence
+
+Question: "Does deleting this adapter break the HTTP client?"
+
+Evidence: imports show no production consumer, tests cover the replacement
+adapter, and history confirms the old adapter was retained only for migration.
+After checking dynamic loading and configuration references, report at level
+4-5. Cite each source and distinguish observed facts from the final inference.
+
+### Cross-Module Finding, Incomplete Evidence
+
+Question: "Why do all write operations depend on the legacy package?"
+
+Evidence: searches show imports in several modules, but runtime registration
+and generated code have not been inspected. Stay at level 2-3. Continue tracing
+composition roots and tests before claiming an architectural dependency.
+
+### Early Delivery
+
+The user asks for findings before runtime behavior or history can be checked.
+Deliver the verified portion at its current level and include `△ Caveats` with:
+
+- the evidence already inspected;
+- assumptions required by the conclusion;
+- unverified modules or execution paths;
+- the next checks that would materially change confidence.
+
+Do not present the early report as level 5 even when the likely explanation is
+simple.
+
+## Delivery Check
+
+Before assigning level 4 or 5, confirm that:
+
+- more than one relevant evidence source supports the conclusion;
+- direct observations and inferences are labeled separately;
+- at least one plausible counter-example was considered;
+- the investigated scope matches the wording of the conclusion;
+- remaining uncertainty is absent or explicitly bounded.
.gitea/ci/thirdparty_skills.json
+33 -38
View File
@@ -13,35 +13,25 @@
"skills/misc"
],
"include_skill_dirs": [
"ask-matt",
"code-review",
"codebase-design",
"diagnosing-bugs",
"domain-modeling",
"git-guardrails-claude-code",
"grill-me",
"grill-with-docs",
"grilling",
"handoff",
"implement",
"improve-codebase-architecture",
"migrate-to-shoehorn",
"prototype",
"research",
"resolving-merge-conflicts",
"scaffold-exercises",
"setup-matt-pocock-skills",
"setup-pre-commit",
"tdd",
"teach",
"to-questionnaire",
"to-spec",
"to-tickets",
"triage",
"wait-what",
"wayfinder",
"wizard",
"writing-for-agents"
"wizard"
]
},
{
@@ -57,15 +47,6 @@
"data_dir": "src/ui-ux-pro-max/data",
"scripts_dir": "src/ui-ux-pro-max/scripts"
},
{
"id": "andrej-karpathy-skills",
"upstream_repo": "https://github.com/forrestchang/andrej-karpathy-skills.git",
"upstream_ref": "main",
"snapshot_dir": "andrej-karpathy-skills",
"sync_mode": "copy_skill_dirs",
"source_list": "skills/thirdparty/.sources/andrej-karpathy-skills.list",
"skills_subdir": "skills"
},
{
"id": "brooks-lint",
"upstream_repo": "https://github.com/hyhmrright/brooks-lint.git",
@@ -92,27 +73,41 @@
"sync_mode": "copy_skill_dirs",
"source_list": "skills/thirdparty/.sources/codebase-recon.list",
"skills_subdir": "plugins/outfitter/skills",
"include_skill_dirs": ["codebase-recon", "pathfinding"]
"include_skill_dirs": ["codebase-recon"],
"overlay_patch": ".gitea/ci/thirdparty-skill-overlays/codebase-recon.patch"
},
{
"id": "codebase-migrate",
"upstream_repo": "https://github.com/ComposioHQ/awesome-codex-skills.git",
"upstream_ref": "master",
"snapshot_dir": "awesome-codex-skills",
"sync_mode": "copy_skill_dirs",
"source_list": "skills/thirdparty/.sources/codebase-migrate.list",
"skills_subdir": ".",
"include_skill_dirs": ["codebase-migrate"]
},
{
"id": "uncle-bob-craft",
"upstream_repo": "https://github.com/sickn33/antigravity-awesome-skills.git",
"id": "cangjie-skill",
"upstream_repo": "https://github.com/kangarooking/cangjie-skill.git",
"upstream_ref": "main",
"snapshot_dir": "antigravity-awesome-skills",
"sync_mode": "copy_skill_dirs",
"source_list": "skills/thirdparty/.sources/uncle-bob-craft.list",
"skills_subdir": "skills",
"include_skill_dirs": ["uncle-bob-craft"]
"snapshot_dir": "cangjie-skill",
"sync_mode": "copy_skill_root",
"source_list": "skills/thirdparty/.sources/cangjie-skill.list",
"output_name": "cangjie-skill",
"include_paths": [
"SKILL.md",
"LICENSE",
"methodology",
"extractors",
"templates"
]
},
{
"id": "darwin-skill",
"upstream_repo": "https://github.com/alchaincyf/darwin-skill.git",
"upstream_ref": "master",
"snapshot_dir": "darwin-skill",
"sync_mode": "copy_skill_root",
"source_list": "skills/thirdparty/.sources/darwin-skill.list",
"output_name": "darwin-skill",
"include_paths": [
"SKILL.md",
"LICENSE",
"references",
"scripts",
"templates",
"test-prompts.json"
]
}
]
}
.gitea/workflows/update-thirdparty-skills.yml
+153 -5
View File
@@ -293,7 +293,26 @@ jobs:
exit 1
fi
# Remove snapshot directories that were managed by an older manifest.
# Managed snapshots are identified by their root SOURCE.md marker.
declare -A active_snapshots=()
while IFS=$'\x1f' read -r source_id upstream_repo upstream_ref snapshot_dir sync_mode remove_paths; do
[ -n "$source_id" ] || continue
active_snapshots["$snapshot_dir"]=1
done < "$sources_file"
changed=0
for source_marker in */SOURCE.md; do
[ -f "$source_marker" ] || continue
stale_snapshot="${source_marker%/SOURCE.md}"
if [ -z "${active_snapshots[$stale_snapshot]:-}" ]; then
echo "Removing orphaned third-party snapshot: $stale_snapshot"
rm -rf -- "$stale_snapshot"
git add -A -- "$stale_snapshot"
changed=1
fi
done
while IFS=$'\x1f' read -r source_id upstream_repo upstream_ref snapshot_dir sync_mode remove_paths; do
[ -n "$source_id" ] || continue
remove_paths_md="${remove_paths//$'\x1e'/,}"
@@ -377,8 +396,9 @@ jobs:
echo "📌 Current $THIRDPARTY_BRANCH: $after_ref"
if [ "$after_ref" = "$before_ref" ]; then
echo "✅ No thirdparty snapshot change; skip main sync."
exit 0
echo "✅ No thirdparty snapshot change; continue with manifest-driven main sync."
else
echo "📌 Updated $THIRDPARTY_BRANCH: $before_ref -> $after_ref"
fi
echo "========================================"
@@ -428,6 +448,8 @@ jobs:
entry.get("data_dir", ""),
entry.get("scripts_dir", ""),
"\x1e".join(entry.get("include_skill_dirs", [])),
entry.get("overlay_patch", ""),
"\x1e".join(entry.get("include_paths", [])),
]
)
)
@@ -521,6 +543,96 @@ jobs:
return 1
}
apply_overlay_patch() {
local source_id="$1"
local patch_file="$2"
[ -n "$patch_file" ] || return 0
if [ ! -f "$patch_file" ]; then
echo "ERROR: overlay patch not found for $source_id: $patch_file" >&2
exit 1
fi
if ! git apply --check "$patch_file"; then
echo "ERROR: overlay patch no longer applies for $source_id: $patch_file" >&2
exit 1
fi
git apply "$patch_file"
echo "Applied overlay patch for $source_id: $patch_file"
}
copy_skill_root() {
local snapshot_root="$1"
local output_dir="$2"
local include_paths="$3"
python3 - "$snapshot_root" "$output_dir" "$include_paths" <<'PY'
import os
import pathlib
import shutil
import sys
snapshot_root = pathlib.Path(sys.argv[1]).resolve(strict=True)
output_dir = pathlib.Path(sys.argv[2])
raw_paths = [value for value in sys.argv[3].split("\x1e") if value]
if not raw_paths:
raise SystemExit("ERROR: copy_skill_root requires include_paths")
sources = []
seen = set()
for value in raw_paths:
relative = pathlib.PurePosixPath(value)
if (
relative.is_absolute()
or not relative.parts
or any(part in ("", ".", "..") for part in relative.parts)
):
raise SystemExit(f"ERROR: unsafe copy_skill_root include path: {value}")
if value in seen:
raise SystemExit(f"ERROR: duplicate copy_skill_root include path: {value}")
seen.add(value)
source_path = snapshot_root / pathlib.Path(*relative.parts)
if not source_path.exists():
raise SystemExit(
f"ERROR: copy_skill_root include path not found: {value}"
)
source = source_path.resolve(strict=True)
try:
source.relative_to(snapshot_root)
except ValueError:
raise SystemExit(
f"ERROR: copy_skill_root include path escapes snapshot: {value}"
)
if not source.is_file() and not source.is_dir():
raise SystemExit(
f"ERROR: unsupported copy_skill_root source type: {value}"
)
sources.append((relative, source))
if os.path.lexists(output_dir):
if output_dir.is_symlink() or output_dir.is_file():
output_dir.unlink()
else:
shutil.rmtree(output_dir)
output_dir.mkdir(parents=True)
for relative, source in sources:
destination = output_dir.joinpath(*relative.parts)
destination.parent.mkdir(parents=True, exist_ok=True)
if source.is_dir():
shutil.copytree(source, destination)
else:
shutil.copy2(source, destination)
if not (output_dir / "SKILL.md").is_file():
raise SystemExit(
f"ERROR: copy_skill_root output is missing SKILL.md: {output_dir}"
)
PY
}
cd "$REPO_DIR"
@@ -546,19 +658,39 @@ jobs:
exit 1
fi
while IFS=$'\x1f' read -r source_id snapshot_dir sync_mode source_list skills_subdirs output_name platform_config template_root data_dir scripts_dir include_skill_dirs; do
# Clean skill directories and source lists left by sources removed from
# the manifest. Active lists are cleared below before being rewritten.
declare -A active_source_lists=()
while IFS=$'\x1f' read -r source_id snapshot_dir sync_mode source_list skills_subdirs output_name platform_config template_root data_dir scripts_dir include_skill_dirs overlay_patch include_paths; do
[ -n "$source_id" ] || continue
active_source_lists["$source_list"]=1
done < "$sources_file"
for source_list in skills/thirdparty/.sources/*.list; do
[ -f "$source_list" ] || continue
if [ -z "${active_source_lists[$source_list]:-}" ]; then
echo "Removing orphaned third-party source list: $source_list"
while IFS= read -r name; do
[ -n "$name" ] || continue
rm -rf -- "skills/thirdparty/$name"
done < "$source_list"
rm -f -- "$source_list"
fi
done
while IFS=$'\x1f' read -r source_id snapshot_dir sync_mode source_list skills_subdirs output_name platform_config template_root data_dir scripts_dir include_skill_dirs overlay_patch include_paths; do
[ -n "$source_id" ] || continue
if [ -f "$source_list" ]; then
while IFS= read -r name; do
[ -n "$name" ] || continue
rm -rf "skills/$name"
rm -rf "skills/thirdparty/$name"
done < "$source_list"
fi
done < "$sources_file"
declare -A owners=()
while IFS=$'\x1f' read -r source_id snapshot_dir sync_mode source_list skills_subdirs output_name platform_config template_root data_dir scripts_dir include_skill_dirs; do
while IFS=$'\x1f' read -r source_id snapshot_dir sync_mode source_list skills_subdirs output_name platform_config template_root data_dir scripts_dir include_skill_dirs overlay_patch include_paths; do
[ -n "$source_id" ] || continue
git archive --format=tar "origin/${THIRDPARTY_BRANCH}" "$snapshot_dir" | tar -xf - -C "$tmp_dir"
@@ -615,12 +747,28 @@ jobs:
names+=("$name")
owners["$name"]="$source_id"
;;
copy_skill_root)
name="$output_name"
if [ -n "${owners[$name]:-}" ] && [ "${owners[$name]}" != "$source_id" ]; then
echo "ERROR: duplicate third-party skill name: $name" >&2
exit 1
fi
if tracked_skill_exists "$name"; then
echo "ERROR: skill name conflict with tracked skill: $name" >&2
exit 1
fi
copy_skill_root "$snapshot_root" "skills/thirdparty/$name" "$include_paths"
names+=("$name")
owners["$name"]="$source_id"
;;
*)
echo "ERROR: unsupported sync mode: $sync_mode" >&2
exit 1
;;
esac
apply_overlay_patch "$source_id" "$overlay_patch"
printf "%s\n" "${names[@]}" | sort > "$source_list"
done < "$sources_file"
README.md
+5 -5
View File
@@ -110,10 +110,10 @@ TSL 语法事实唯一由 `tsl-syntax-reference` 管理；`rulesets/tsl/index.md
- `rulesets/index.md`：规则集索引（跨语言）
- `rulesets/tsl/index.md`：TSL 领域路由与事实边界
- `rulesets/cpp/index.md`：C++ 核心约定（46 行）
- `rulesets/python/index.md`：Python 核心约定（44 行）
- `rulesets/typescript/index.md`：TypeScript 核心约定（47 行）
- `rulesets/markdown/index.md`：Markdown 核心约定（31 行，仅代码格式化）
- `rulesets/cpp/index.md`：C++ 核心约定
- `rulesets/python/index.md`：Python 核心约定
- `rulesets/typescript/index.md`：TypeScript/JavaScript 核心约定
- `rulesets/markdown/index.md`：Markdown 格式化边界（仅代码格式化）
更多说明：`rulesets/index.md`
@@ -247,7 +247,7 @@ skills = ["tsl-syntax-reference", "tsl-api-reference"]
1. **仓库级（跨语言）共识**：对所有语言都成立的规则与流程。
- 提交信息：`docs/common/commit_message.md`
- 行尾与文本规范：`.gitattributes`
- 智能体最低要求：`.agents/*`（工作原则、质量底线、安全边界）
- 智能体最低要求：`.agents/*`（语言硬约束、事实所有者路由和阻断边界）
2. **语言级（Language-specific）规范**：只对某个语言成立的风格与工具。
- 例如 TSL 的命名/文件顶层声明限制、C++ 的 `.clang-format/.clang-tidy`、Python 的 `ruff`、TypeScript 的 ESLint/类型约束等。
SKILLS.md
+2 -1
View File
@@ -171,7 +171,8 @@ skills/
**来源登记**：
- `skills/thirdparty/.sources/`：来源清单目录（`*.list` 文件）
- `skills/thirdparty/thirdparty-skills.yml`：统一配置文件
- `.gitea/ci/thirdparty_skills.json`：工作流读取的同步来源清单
- `skills/thirdparty/thirdparty-skills.yml`：供人工查阅的第三方来源登记表
**同步流程**：
rulesets/cpp/index.md
-7
View File
@@ -7,13 +7,6 @@
- 作为仓库级基线规则集使用；更靠近代码目录的规则更具体并可覆盖基线。
- 当智能体规则与 docs 冲突：安全/合规优先，其次保持仓库一致性。
## 智能体工作原则（铁律）
1. 先理解目标与上下文，再动手改代码
2. 修改要小而清晰；避免无关重构
3. 发现安全问题（内存安全/鉴权漏洞）立即标注或修复
4. 不引入新依赖或工具，除非明确要求
## C++ 核心约定（不可违反）
- 使用 C++23，关闭编译器扩展，并固定经过验证的 CMake、编译器和 Conan 2 版本。
rulesets/markdown/index.md
+1 -1
View File
@@ -2,7 +2,7 @@
本规则集定义 AI/自动化智能体在处理 Markdown（`.md`）文件时必须遵守的核心约束。
## 智能体工作原则（铁律）
## Markdown 格式化边界
1. 只调整代码块与行内代码；不改写正文内容
2. 不改变标题层级、列表结构、段落顺序
rulesets/python/index.md
-7
View File
@@ -7,13 +7,6 @@
- 作为仓库级基线规则集使用；更靠近代码目录的规则更具体并可覆盖基线。
- 当智能体规则与 docs 冲突：安全/合规优先，其次保持仓库一致性。
## 智能体工作原则（铁律）
1. 先理解目标与上下文，再动手改代码
2. 修改要小而清晰；避免无关重构
3. 发现安全问题（明文密钥/鉴权漏洞）立即标注或修复
4. 不引入新依赖或工具，除非明确要求
## Python 核心约定（不可违反）
- 代码风格：Google Python Style Guide；优先使用仓库既有配置（`pyproject.toml`, `.flake8`, `.pylintrc`）
rulesets/typescript/index.md
-7
View File
@@ -7,13 +7,6 @@
- 作为仓库级基线规则集使用；更靠近代码目录的规则更具体并可覆盖基线。
- 当智能体规则与 docs 冲突：安全/合规优先，其次保持仓库一致性。
## 智能体工作原则（铁律）
1. 先理解目标与上下文，再动手改代码
2. 修改要小而清晰；避免无关重构
3. 发现安全问题（明文密钥/鉴权漏洞）立即标注或修复
4. 不引入新依赖或工具，除非明确要求
## TypeScript/JavaScript 核心约定（不可违反）
- 语言标准：业务代码优先 TypeScript（`.ts/.tsx`）；JavaScript（`.js/.mjs/.cjs`）仅用于脚本、配置或兼容场景
skills/README.md
+9 -11
View File
@@ -28,7 +28,7 @@
| `grill-with-docs` / `grilling` | 收敛 design tree 并同步领域词汇/ADR |
| `to-spec` | 把已确认上下文合成为 feature spec |
| `to-tickets` | 拆成带 blocking edges 的 tracer-bullet tickets |
| `implement` / `tdd` | 在约定 seam 上按 ticket 实现和验证 |
| `tdd` | 在约定 seam 上按 ticket 采用测试驱动实现和验证 |
| `code-review` | 对固定点后的提交做 Standards/Spec 双轴审查 |
| `handoff` | 跨 session 交接未完成上下文 |
@@ -51,16 +51,14 @@ Code review 和质量审查工具集，包含共享参考资料目录 `_shared`
### 代码库侦察与迁移
| Skill | 作用 |
| ------------------ | --------------------------------------------------------- |
| `codebase-recon` | 通过 git 历史和代码结构做 risk scan、热点分析、重构前侦察 |
| `pathfinding` | 需求澄清和 confidence calibration |
| `codebase-migrate` | 大代码库迁移、多文件 refactor、分批变更与 CI 验证工作流 |
| Skill | 作用 |
| ---------------- | --------------------------------------------------------- |
| `codebase-recon` | 通过 git 历史和代码结构做 risk scan、热点分析、重构前侦察 |
### 其他
| Skill | 作用 |
| --------------------- | ------------------------------------------------------------- |
| `ui-ux-pro-max` | UI/UX 设计知识库，包含风格、配色、字体、组件和栈相关建议 |
| `karpathy-guidelines` | 写代码、审查、重构时降低 LLM 常见错误：保持简单、聚焦、可验证 |
| `uncle-bob-craft` | Clean Architecture、SOLID、设计模式误用和代码工匠实践审查 |
| Skill | 作用 |
| --------------- | -------------------------------------------------------- |
| `ui-ux-pro-max` | UI/UX 设计知识库，包含风格、配色、字体、组件和栈相关建议 |
| `cangjie-skill` | 将书籍、课程、访谈等长内容蒸馏为原子化、可执行的 skills |
| `darwin-skill` | 评估并迭代优化 skills，以测试提示和成对评审控制质量回归 |
skills/thirdparty/thirdparty-skills.yml
Vendored
+38 -29
View File
@@ -31,46 +31,55 @@ skills:
upstream_ref: main
upstream_paths:
- plugins/outfitter/skills/codebase-recon
- plugins/outfitter/skills/pathfinding
upstream_layout: multi_skill_subset
selected_for:
- risk scan
- git-history hotspot analysis
- codebase reconnaissance before planning
- refactor-risk assessment
- confidence calibration reference dependency
- evidence-based confidence calibration
playbook_fit: pre-design and pre-refactor codebase risk discovery
notes:
- Upstream skills live inside a larger agents repository.
- pathfinding is included because codebase-recon links to pathfinding/references/confidence.md.
- A sync overlay makes codebase-recon confidence calibration self-contained.
- id: codebase-migrate
- id: cangjie-skill
sync: enabled
upstream_repo: https://github.com/ComposioHQ/awesome-codex-skills
upstream_ref: master
upstream_path: codebase-migrate
upstream_layout: single_skill_dir
selected_for:
- large codebase migrations
- multi-file refactors
- reviewable refactor batches
- CI-verified migration workflows
notes:
- Sourced from awesome-codex-skills.
- id: uncle-bob-craft
sync: enabled
upstream_repo: https://github.com/sickn33/antigravity-awesome-skills
upstream_repo: https://github.com/kangarooking/cangjie-skill
upstream_ref: main
upstream_path: skills/uncle-bob-craft
upstream_layout: single_skill_from_suite
upstream_layout: repository_root_skill
upstream_paths:
- SKILL.md
- LICENSE
- methodology
- extractors
- templates
selected_for:
- Clean Architecture review
- SOLID review
- design-pattern misuse checks
- architecture-boundary refactoring
- code craftsmanship review
playbook_fit: architecture and code-quality craft review support
- distilling books and long-form content into executable skills
- extracting reusable frameworks, principles, and methods
- generating Darwin-compatible skill test prompts
playbook_fit: long-form knowledge distillation into reusable skills
notes:
- Upstream skill includes references for Clean Architecture, Clean Coder, Clean Agile, and design-pattern discipline.
- Registered as a single selected skill from a larger community skill suite.
- The repository root is the skill directory.
- Licensed under AGPL-3.0.
- id: darwin-skill
sync: enabled
upstream_repo: https://github.com/alchaincyf/darwin-skill
upstream_ref: master
upstream_layout: repository_root_skill
upstream_paths:
- SKILL.md
- LICENSE
- references
- scripts
- templates
- test-prompts.json
selected_for:
- skill quality assessment
- test-prompt-driven skill optimization
- paired evaluation and regression control
playbook_fit: evaluation and iterative optimization of skills
notes:
- The repository root is the skill directory.
- Licensed under MIT.
templates/AGENT_RULES.template.md
+6 -9
View File
@@ -80,7 +80,7 @@ setup-matt-pocock-skills
```
- 每个仓库首次使用时运行 `setup-matt-pocock-skills`，本地开发选择 local markdown tracker
- 正式仓库设计使用 `grill-with-docs`；`grill-me` 只用于无仓库的一次性讨论
- 正式仓库设计使用 `grill-with-docs`
- 进入 `grill-with-docs` 或本地 ticket 执行协议前，重新读取 `memory-bank/project-brief.md`、
`memory-bank/tech-context.md` 和 `memory-bank/system-patterns.md`
- `grilling` 必须走完整 design tree，frontier 清空并经用户确认后才进入 `to-spec`
@@ -96,8 +96,6 @@ setup-matt-pocock-skills
`to-spec -> to-tickets`，不得从决策 ticket 直接跳到实现
- `research` 产出的高可信一手来源报告先进入 `grill-with-docs`，作为设计输入；调研不能
替代 grilling
- 设计问题需要 runnable answer 时，从原设计会话 `handoff` 到独立目录运行
`prototype`，再 `handoff` 结论回原设计会话
- 已由 `to-tickets` 生成的 ticket 直接从 `main_loop.py claim -> 本地 ticket 执行协议` 开始，
不再 triage 或重复需求采访
@@ -367,12 +365,11 @@ partial feature 还需 `--allow-partial`。返回 `RETRY: feature needs main syn
以下能力在对应阶段内按需使用，不替代上面的入口路由和证据门禁：
| Skill | 触发条件 |
| ------------------ | ----------------------------------- |
| `codebase-recon` | 架构、跨模块、重构、迁移或风险不明 |
| `brooks-audit` | 架构边界和长期维护性审查 |
| `codebase-migrate` | 大规模迁移或宽重构 |
| `commit-message` | 需要检查 staged diff 或生成提交信息 |
| Skill | 触发条件 |
| ---------------- | ----------------------------------- |
| `codebase-recon` | 架构、跨模块、重构、迁移或风险不明 |
| `brooks-audit` | 架构边界和长期维护性审查 |
| `commit-message` | 需要检查 staged diff 或生成提交信息 |
## 需要确认的场景
templates/README.md
+1 -1
View File
@@ -108,7 +108,7 @@ setup-matt-pocock-skills
-> to-tickets
-> main_loop.py enqueue
-> main_loop.py claim
-> implement + tdd
-> 本地 ticket 执行协议（tdd）
-> commit
-> code-review
-> main_loop.py finish