Agent规划
- 作者仓库星标 435
- 作者更新于 实时读取
- 作者仓库 nvim
- 领域
- 工程开发
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 88 / 100 · 社区维护
- 作者 / 版本 / 许可
- @benbrastmckie · 未声明 license
- Token 消耗评级
- 中等消耗
- 接入复杂程度
- 需简单配置
- 是否需要外部 API Key
- 不需要
- 兼容的系统
- 未声明(默认跨平台)
- 底层运行要求
- 无特殊要求
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- Shell 执行
- 网络行为
- 仅限本地
- 安装命令数
- 26 条
档案由构建时根据 SKILL.md 与安装命令自动衍生,可能与作者实际意图存在差异。
需要注意: 未限定 allowed-tools,默认拥有全部工具权限。
---
name: skill-web-implementation
description: Implement web (Astro/Tailwind/TypeScript) changes following a plan. Invoke for web-language impl…
category: 工程开发
runtime: 无特殊运行时
---
# skill-web-implementation 输出预览
## PART A: 任务判断
- 适用问题:代码实现、重构、调试或代码审查。
- 输入要求:目标材料、限制条件、期望输出和验收方式。
- 证据边界:围绕“Context References / Trigger Conditions / Execution”读取原文规则,不把推断写成作者承诺。
## PART B: 执行结果
- **01** 任务判断:确认你的需求是否属于代码实现、重构、调试或代码审查,并标出输入、限制和预期结果。
- **02** 执行计划:优先按“Context References / Trigger Conditions / Execution”拆成步骤,说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果:给出可复制的命令、文件改动、检查清单或内容草稿,并说明如何继续迭代。
- **04** 风险边界:结合 读取文件、写入/修改文件、执行终端命令、主要在本地完成、通常不需要额外 API Key 给出执行前确认项。
## Running Rules
- 读取文件、写入/修改文件、执行终端命令;主要在本地完成;通常不需要额外 API Key。
- 先小样例验证,再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。 原文出现了 `/implement` 这类斜杠命令;如果你的 Agent 支持命令触发,优先用命令开场,再补充目标和边界。
告诉 Agent 目标文件或材料、期望结果、不可改范围、是否允许联网或执行命令。本 Skill 的权限画像是:读取文件、写入/修改文件、执行终端命令。
先用一个小任务确认它会围绕“Context References / Trigger Conditions / Execution”工作;涉及文件或命令时,先看 diff、日志、预览或测试结果。
检查最终产物是否包含明确结果、必要证据和下一步动作;如果输出泛泛而谈,就补充输入、边界和验收标准后重跑。
---
name: skill-web-implementation
description: Implement web (Astro/Tailwind/TypeScript) changes following a plan. Invoke for web-language impl…
category: 工程开发
source: benbrastmckie/nvim
---
# skill-web-implementation
## 什么时候使用
- 把工程方向的常用动作沉淀成 Agent 可调用的技能 适合处理工程开发场景下的代码实现、调试、重构、测试或代码审查,核心价值是把输入、判断、执行、验证和交付边界固定下来,避免 Agent 泛泛回答。 把任务拆成可执行、可检查、可继续迭代…
- 面向代码实现、重构、调试或代码审查,优先处理能明确输入、步骤和验收标准的工作。
## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。
## 执行规则
- 围绕「Context References / Trigger Conditions / Execution」组织步骤,不把推断写成作者事实。
- 读取文件、写入/修改文件、执行终端命令;主要在本地完成;通常不需要额外 API Key。
- 先跑小样例,确认结果可检查后再扩大任务范围。
## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown,不编造命令、平台或依赖。 作者原文负责流程事实;仓库文件负责来源和命令;流狐只补充适用场景、限制和质量判断。
skill "skill-web-implementation" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> Context References / Trigger Conditions / Execution
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> 无特殊运行时 | 读取文件、写入/修改文件、执行终端命令 | 主要在本地完成
安全层 -> 通常不需要额外 API Key + 小任务验证 + diff / 日志复核
输出层 -> 可复制结果 + 检查清单 + 下一步迭代
} Web Implementation Skill
Thin wrapper that delegates web (Astro/Tailwind/TypeScript) implementation to web-implementation-agent subagent.
IMPORTANT: This skill implements the skill-internal postflight pattern. After the subagent returns, this skill handles all postflight operations (status update, artifact linking, git commit) before returning. This eliminates the "continue" prompt issue between skill return and orchestrator.
Context References
Reference (do not load eagerly):
- Path:
.claude/context/project/web/README.md- Web context overview - Path:
.claude/context/project/web/domain/astro-framework.md- Astro reference - Path:
.claude/context/project/web/domain/tailwind-v4.md- Tailwind v4 reference
Note: This skill is a thin wrapper with internal postflight. Context is loaded by the delegated agent.
Trigger Conditions
This skill activates when:
- Task type is "web"
- /implement command targets a web (Astro/Tailwind/TypeScript) task
- Pages, components, layouts, or web styling needs to be created or modified
Execution
0. Preflight Status Update
Before delegating to the subagent, update task status to "implementing".
Update state.json:
jq --arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \
--arg status "implementing" \
--arg sid "$session_id" \
'(.active_projects[] | select(.project_number == '$task_number')) |= . + {
status: $status,
last_updated: $ts,
session_id: $sid,
started: $ts
}' specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json
Update TODO.md: Use Edit tool to change status marker from [PLANNED] to [IMPLEMENTING].
Update plan file (if exists): Update the Status field in plan metadata:
# Find latest plan file
plan_file=$(ls -1 "specs/${padded_num}_${project_name}/plans/"*.md 2>/dev/null | sort -V | tail -1)
if [ -n "$plan_file" ] && [ -f "$plan_file" ]; then
sed -i "s/^\- \*\*Status\*\*: \[.*\]$/- **Status**: [IMPLEMENTING]/" "$plan_file"
fi
Create Postflight Marker:
# Ensure task directory exists
padded_num=$(printf "%03d" "$task_number")
mkdir -p "specs/${padded_num}_${project_name}"
cat > "specs/${padded_num}_${project_name}/.postflight-pending" << EOF
{
"session_id": "${session_id}",
"skill": "skill-web-implementation",
"task_number": ${task_number},
"operation": "implement",
"reason": "Postflight pending: status update, artifact linking, git commit",
"created": "$(date -u +%Y-%m-%dT%H:%M:%SZ)",
"stop_hook_active": false
}
EOF
1. Input Validation
Validate required inputs:
task_number- Must be provided and exist in state.json- Task status must allow implementation (planned, implementing, partial)
# Lookup task
task_data=$(jq -r --arg num "$task_number" \
'.active_projects[] | select(.project_number == ($num | tonumber))' \
specs/state.json)
# Validate exists
if [ -z "$task_data" ]; then
return error "Task $task_number not found"
fi
# Extract fields
task_type=$(echo "$task_data" | jq -r '.task_type // "general"')
status=$(echo "$task_data" | jq -r '.status')
project_name=$(echo "$task_data" | jq -r '.project_name')
description=$(echo "$task_data" | jq -r '.description // ""')
# Validate language
if [ "$task_type" != "web" ]; then
return error "Task $task_number is not a web task"
fi
# Validate status
if [ "$status" = "completed" ]; then
return error "Task already completed"
fi
2. Context Preparation
Prepare delegation context:
{
"session_id": "sess_{timestamp}_{random}",
"delegation_depth": 1,
"delegation_path": ["orchestrator", "implement", "skill-web-implementation"],
"timeout": 3600,
"task_context": {
"task_number": N,
"task_name": "{project_name}",
"description": "{description}",
"task_type": "web"
},
"plan_path": "specs/{NNN}_{SLUG}/plans/MM_{short-slug}.md",
"metadata_file_path": "specs/{NNN}_{SLUG}/.return-meta.json"
}
3. Invoke Subagent
CRITICAL: You MUST use the Agent tool to spawn the subagent.
The agent field in this skill's frontmatter specifies the target: web-implementation-agent
Required Tool Invocation:
Tool: Agent (NOT Skill, NOT Plan)
Parameters:
- subagent_type: "web-implementation-agent"
- prompt: [Include task_context, delegation_context, plan_path]
- description: "Execute web implementation for task {N}"
DO NOT use Skill(web-implementation-agent) - this will FAIL.
Agents live in .claude/agents/, not .claude/skills/.
The Skill tool can only invoke skills from .claude/skills/.
The subagent will:
- Load web-specific context files (Astro framework, Tailwind v4, style guide, etc.)
- Create/modify .astro, .ts, .tsx, .css files
- Execute build verification (pnpm build, pnpm check)
- Handle TypeScript and Astro errors
- Create implementation summary
- Write metadata to
specs/{NNN}_{SLUG}/.return-meta.json - Return a brief text summary (NOT JSON)
3a. Validate Subagent Return Format
IMPORTANT: Check if subagent accidentally returned JSON to console (v1 pattern) instead of writing to file (v2 pattern).
If the subagent's text return parses as valid JSON, log a warning:
# Check if subagent return looks like JSON (starts with { and is valid JSON)
subagent_return="$SUBAGENT_TEXT_RETURN"
if echo "$subagent_return" | grep -q '^{' && echo "$subagent_return" | jq empty 2>/dev/null; then
echo "WARNING: Subagent returned JSON to console instead of writing metadata file."
echo "This indicates the agent may have outdated instructions (v1 pattern instead of v2)."
echo "The skill will continue by reading the metadata file, but this should be fixed."
fi
This validation:
- Does NOT fail the operation (continues to read metadata file)
- Logs a warning for debugging
- Indicates the subagent instructions need updating
- Allows graceful handling of mixed v1/v2 agents
3b. Self-Execution Fallback
CRITICAL: If you performed the work above WITHOUT using the Agent tool (i.e., you read files,
wrote artifacts, or updated metadata directly instead of spawning a subagent), you MUST write a
.return-meta.json file now before proceeding to postflight. Use the schema from
return-metadata-file.md with status value "implemented".
If you DID use the Agent tool, skip this stage -- the subagent already wrote the metadata.
Postflight (ALWAYS EXECUTE)
The following stages MUST execute after work is complete, whether the work was done by a subagent or inline (Stage 3b). Do NOT skip these stages for any reason.
4. Parse Subagent Return (Read Metadata File)
Read the metadata file:
metadata_file="specs/${padded_num}_${project_name}/.return-meta.json"
if [ -f "$metadata_file" ] && jq empty "$metadata_file" 2>/dev/null; then
status=$(jq -r '.status' "$metadata_file")
artifact_path=$(jq -r '.artifacts[0].path // ""' "$metadata_file")
artifact_type=$(jq -r '.artifacts[0].type // ""' "$metadata_file")
artifact_summary=$(jq -r '.artifacts[0].summary // ""' "$metadata_file")
phases_completed=$(jq -r '.metadata.phases_completed // 0' "$metadata_file")
phases_total=$(jq -r '.metadata.phases_total // 0' "$metadata_file")
# Extract completion_data fields (if present)
completion_summary=$(jq -r '.completion_data.completion_summary // ""' "$metadata_file")
roadmap_items=$(jq -c '.completion_data.roadmap_items // []' "$metadata_file")
else
echo "Error: Invalid or missing metadata file"
status="failed"
fi
Validate the metadata contains required fields:
- Status is one of: implemented, partial, failed, blocked
- Summary is non-empty and <100 tokens
- Artifacts array present (source files, summary)
- Metadata contains session_id, agent_type, delegation info
5. Postflight Status Update
After implementation, update task status based on result.
If result.status == "implemented":
Update state.json to "completed" and add completion_data fields (two-step pattern):
# Step 1: Update status and timestamps
jq --arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \
--arg status "completed" \
'(.active_projects[] | select(.project_number == '$task_number')) |= . + {
status: $status,
last_updated: $ts,
completed: $ts
}' specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json
# Step 2: Add completion_summary (always required for completed tasks)
if [ -n "$completion_summary" ]; then
jq --arg summary "$completion_summary" \
'(.active_projects[] | select(.project_number == '$task_number')).completion_summary = $summary' \
specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json
fi
# Step 3: Add roadmap_items (if present and non-empty)
if [ "$roadmap_items" != "[]" ] && [ -n "$roadmap_items" ]; then
jq --argjson items "$roadmap_items" \
'(.active_projects[] | select(.project_number == '$task_number')).roadmap_items = $items' \
specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json
fi
# Step 4: Filter out existing summary artifacts (use "| not" pattern to avoid != escaping)
jq '(.active_projects[] | select(.project_number == '$task_number')).artifacts =
[(.active_projects[] | select(.project_number == '$task_number')).artifacts // [] | .[] | select(.type == "summary" | not)]' \
specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json
# Step 5: Add new summary artifact
jq --arg path "$artifact_path" \
'(.active_projects[] | select(.project_number == '$task_number')).artifacts += [{"path": $path, "type": "summary"}]' \
specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json
Update TODO.md:
- Change status marker from
[IMPLEMENTING]to[COMPLETED] - Link artifact using count-aware format: apply the four-case Edit logic from
@.claude/context/patterns/artifact-linking-todo.mdwithfield_name=**Summary**,next_field=**Description**
Update plan file (if exists): Update the Status field to [COMPLETED]:
plan_file=$(ls -1 "specs/${padded_num}_${project_name}/plans/"*.md 2>/dev/null | sort -V | tail -1)
if [ -n "$plan_file" ] && [ -f "$plan_file" ]; then
sed -i "s/^\- \*\*Status\*\*: \[.*\]$/- **Status**: [COMPLETED]/" "$plan_file"
fi
If result.status == "partial":
Update state.json with resume point (keep status as "implementing"):
jq --arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \
--arg phase "$completed_phase" \
'(.active_projects[] | select(.project_number == '$task_number')) |= . + {
last_updated: $ts,
resume_phase: ($phase | tonumber + 1)
}' specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json
TODO.md stays as [IMPLEMENTING].
Update plan file (if exists): Update the Status field to [PARTIAL]:
plan_file=$(ls -1 "specs/${padded_num}_${project_name}/plans/"*.md 2>/dev/null | sort -V | tail -1)
if [ -n "$plan_file" ] && [ -f "$plan_file" ]; then
sed -i "s/^\- \*\*Status\*\*: \[.*\]$/- **Status**: [PARTIAL]/" "$plan_file"
fi
On failed: Do NOT run postflight. Keep status as "implementing" for retry. Do not update plan file (leave as [IMPLEMENTING] for retry).
6. Git Commit
Commit changes with session ID:
git add -A
git commit -m "task ${task_number}: complete implementation
Session: ${session_id}
7. Cleanup
Remove marker and metadata files after postflight processing:
rm -f "specs/${padded_num}_${project_name}/.postflight-pending"
rm -f "specs/${padded_num}_${project_name}/.postflight-loop-guard"
rm -f "specs/${padded_num}_${project_name}/.return-meta.json"
8. Return Brief Summary
Return a brief text summary (NOT JSON) describing the implementation results.
Return Format
This skill returns a brief text summary (NOT JSON). The JSON metadata is written to the file and processed internally.
Example successful return:
Web implementation completed for task 10:
- All 3 phases executed, build passes cleanly
- Created about page with hero section and team grid
- Created summary at specs/10_create_about_page/summaries/01_web-feature-summary.md
- Status updated to [COMPLETED]
- Changes committed with session sess_1770319142_a293c5
Example partial return:
Web implementation partially completed for task 10:
- Phases 1-2 of 3 executed
- Phase 3 blocked: TypeScript error in ContactForm component
- Partial summary at specs/10_create_about_page/summaries/01_web-feature-summary.md
- Status remains [IMPLEMENTING] - run /implement 10 to resume
Error Handling
Input Validation Errors
Return immediately with failed status if task not found, wrong language, or status invalid.
Subagent Errors
Pass through the subagent's error return verbatim.
Timeout
Return partial status if subagent times out (default 3600s).
MUST NOT (Postflight Boundary)
After the agent returns, this skill MUST NOT:
- Edit .astro/.tsx/.ts files - All web implementation is done by agent
- Run pnpm build/check - Build verification is done by agent
- Analyze or grep source - Analysis is agent work
- Write summary/reports - Artifact creation is agent work
PROHIBITION: If the subagent returned partial or failed status, the lead skill MUST NOT attempt to continue, complete, or "fill in" the subagent's work. Report the partial/failed status and let the user re-run
/implementto resume.
The postflight phase is LIMITED TO:
- Reading agent metadata file
- Updating state.json via jq
- Updating TODO.md status marker via Edit
- Linking artifacts in state.json
- Git commit
- Cleanup of temp/marker files
Reference: @.claude/context/standards/postflight-tool-restrictions.md
先判断是否适合
作者设计意图
作者的方法与取舍
边界和复核