Agent 生成器
- 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-founder-implement
description: Execute founder plans and generate strategy reports Routes founder-specific implementation reque…
category: 工程开发
runtime: 无特殊运行时
---
# skill-founder-implement 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 Pointers / Trigger Conditions / Direct Invocation”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 Pointers / Trigger Conditions / Direct Invocation”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`、`/plan` 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 Pointers / Trigger Conditions / Direct Invocation”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-founder-implement
description: Execute founder plans and generate strategy reports Routes founder-specific implementation reque…
category: 工程开发
source: benbrastmckie/nvim
---
# skill-founder-implement
## 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 Pointers / Trigger Conditions / Direct Invocation」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-founder-implement" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> Context Pointers / Trigger Conditions / Direct Invocation
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
} Founder Implement Skill
Routes founder-specific implementation requests to the founder-implement-agent, executing plans created by skill-founder-plan and generating detailed strategy reports.
Context Pointers
Reference (do not load eagerly):
- Path:
.opencode/context/formats/subagent-return.md - Purpose: Return validation
- Load at: Subagent execution only
Note: This skill is a thin wrapper. Context is loaded by the delegated agent, not this skill.
Trigger Conditions
This skill activates when:
Direct Invocation
/implementcommand on a task withlanguage: founder- Extension routing lookup finds
routing.implement.founder
Task-Type-Based Routing
- Task type is "founder"
/implement {N}where task {N} has language="founder"
When NOT to trigger
Do not invoke for:
- Tasks with other language types (general, meta, neovim, etc.)
- Quick mode operations (
--quickflag) - Tasks in [NOT STARTED] status (need plan first)
- Tasks already [COMPLETED]
Execution
1. Input Validation
Validate inputs from delegation context:
task_number- Required, integerplan_path- Required, path to implementation planresume_phase- Optional, phase number to resume fromsession_id- Required, string
# Validate task_number is present
if [ -z "$task_number" ]; then
return error "task_number is required"
fi
# Validate session_id is present
if [ -z "$session_id" ]; then
return error "session_id is required"
fi
# Validate plan_path exists
if [ -z "$plan_path" ] || [ ! -f "$plan_path" ]; then
return error "plan_path is required and must exist. Run /plan first."
fi
2. Preflight Status Update
Update task status to "implementing" in state.json:
jq --argjson num "$task_number" \
--arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \
'(.active_projects[] | select(.project_number == $num)) += {
status: "implementing",
last_updated: $ts
}' specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json
Update TODO.md status marker to [IMPLEMENTING].
3. Create Postflight Marker
Create marker file to signal postflight operations needed:
padded_num=$(printf "%03d" "$task_number")
project_name=$(jq -r --argjson num "$task_number" \
'.active_projects[] | select(.project_number == $num) | .project_name' \
specs/state.json)
task_dir="specs/${padded_num}_${project_name}"
mkdir -p "$task_dir"
cat > "$task_dir/.postflight-pending" << EOF
{
"session_id": "${session_id}",
"skill": "skill-founder-implement",
"task_number": ${task_number},
"operation": "implement",
"reason": "Postflight pending: status update, artifact linking, git commit",
"created": "$(date -u +%Y-%m-%dT%H:%M:%SZ)"
}
EOF
4. Context Preparation
Extract task_type from state.json (null-safe):
# Extract task_type from state.json (null-safe)
task_type=$(jq -r --argjson num "$task_number" \
'.active_projects[] | select(.project_number == $num) | .task_type // null' \
specs/state.json)
Prepare delegation context for agent:
{
"task_context": {
"task_number": 234,
"project_name": "market_sizing_fintech_payments",
"description": "Market sizing: fintech payments",
"task_type": "founder",
"task_type": "market"
},
"plan_path": "specs/234_market_sizing_fintech_payments/plans/01_market-sizing-plan.md",
"resume_phase": 1,
"output_dir": "strategy/",
"metadata_file_path": "specs/234_market_sizing_fintech_payments/.return-meta.json",
"metadata": {
"session_id": "sess_{timestamp}_{random}",
"delegation_depth": 2,
"delegation_path": ["orchestrator", "implement", "skill-founder-implement"]
}
}
5. Invoke Agent
CRITICAL: You MUST use the Task tool to spawn the agent.
Required Tool Invocation:
Tool: Task (NOT Skill)
Parameters:
- subagent_type: "founder-implement-agent"
- prompt: [Include task_context, plan_path, resume_phase, output_dir, metadata]
- description: "Execute founder plan and generate strategy report"
The agent will:
- Load plan and detect resume point
- Execute phases (TAM -> SAM -> SOM -> Report -> Typst/PDF)
- Generate report artifact in
strategy/directory (markdown) - Generate typst/PDF in
founder/directory (if typst installed) - Create summary in task directory
- Write metadata file for postflight consumption
- Return brief text summary
Note: Phase 5 (Typst/PDF generation) is optional. Task completes successfully even if typst is not installed or PDF generation fails.
5b. Self-Execution Fallback
CRITICAL: If you performed the work above WITHOUT using the Task 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 the appropriate status value for this operation.
If you DID use the Task 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 5b). Do NOT skip these stages for any reason.
6. Read Metadata File
Read the metadata file:
metadata_file="specs/${padded_num}_${project_name}/.return-meta.json"
metadata=$(cat "$metadata_file")
status=$(echo "$metadata" | jq -r '.status')
7. Postflight Status Update
If agent succeeded (status == "implemented"):
# Update state.json to completed
jq --argjson num "$task_number" \
--arg ts "$(date -u +%Y-%m-%dT%H:%M:%SZ)" \
'(.active_projects[] | select(.project_number == $num)) += {
status: "completed",
last_updated: $ts
}' specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json
# Read typst metadata fields to determine which artifacts exist
typst_source=$(echo "$metadata" | jq -r '.metadata.typst_source_generated // false')
pdf_compiled=$(echo "$metadata" | jq -r '.metadata.pdf_compiled // false')
# Link artifacts in state.json with typst-primary ordering:
# 1. .typ source file (primary, always generated)
# 2. .pdf compiled output (conditional on pdf_compiled)
# 3. .md markdown report (fallback, always generated)
# 4. summary artifact (always generated)
artifacts=$(echo "$metadata" | jq '.artifacts')
for i in $(seq 0 $(($(echo "$artifacts" | jq 'length') - 1))); do
artifact_type=$(echo "$artifacts" | jq -r ".[$i].type")
artifact_path=$(echo "$artifacts" | jq -r ".[$i].path")
artifact_summary=$(echo "$artifacts" | jq -r ".[$i].summary")
# Skip PDF artifact if pdf_compiled is false
if [ "$pdf_compiled" = "false" ] && echo "$artifact_path" | grep -q '\.pdf$'; then
continue
fi
jq --argjson num "$task_number" \
--arg type "$artifact_type" \
--arg path "$artifact_path" \
--arg summary "$artifact_summary" \
'(.active_projects[] | select(.project_number == $num)).artifacts += [{
type: $type,
path: $path,
summary: $summary
}]' specs/state.json > specs/tmp/state.json && mv specs/tmp/state.json specs/state.json
done
Update TODO.md status marker to [COMPLETED], add Completed date, and link summary artifact per @.opencode/context/patterns/artifact-linking-todo.md with field_name=**Summary**, next_field=**Description**.
If partial (status == "partial"):
- Keep status as "implementing"
- Note resume point in metadata
8. Git Commit
git add -A
git commit -m "$(cat <<'EOF'
task {N}: complete implementation
Session: {session_id}
EOF
)"
9. Cleanup and Return
Remove postflight markers and metadata:
rm -f "$task_dir/.postflight-pending"
rm -f "$task_dir/.postflight-loop-guard"
rm -f "$task_dir/.return-meta.json"
Return brief text summary to caller.
MUST NOT (Postflight Boundary)
After the agent returns -- whether with status implemented, partial, or failed -- this skill MUST proceed immediately to postflight (Stage 6). The skill MUST NOT:
- Edit strategy/report files - All implementation work is done by agent
- Run calculations - Analysis is done by agent
- Use MCP tools - Domain tools are for agent use only
- Analyze or grep source - Analysis is agent work
- Write summary/reports - Artifact creation is done by agent
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: @.opencode/context/standards/postflight-tool-restrictions.md
Return Format
Brief text summary (NOT JSON).
Expected successful return (with PDF compiled):
Founder implementation completed for task {N}:
- Phases {phases_completed}/{phases_total} executed
- TAM: {tam}, SAM: {sam}, SOM Y1: {som_y1}
- Typst source: founder/{report-type}-{slug}.typ (primary)
- PDF report: founder/{report-type}-{slug}.pdf (compiled)
- Markdown: strategy/{report-type}-{slug}.md (fallback)
- Summary: specs/{NNN}_{SLUG}/summaries/01_{short-slug}-summary.md
- Status updated to [COMPLETED]
- Changes committed with session {session_id}
- Next: Review report and validate assumptions
Expected successful return (without PDF - typst CLI not installed):
Founder implementation completed for task {N}:
- Phases {phases_completed}/{phases_total} executed (PDF skipped - typst not installed)
- TAM: {tam}, SAM: {sam}, SOM Y1: {som_y1}
- Typst source: founder/{report-type}-{slug}.typ (primary)
- Markdown: strategy/{report-type}-{slug}.md (fallback)
- Summary: specs/{NNN}_{SLUG}/summaries/01_{short-slug}-summary.md
- Status updated to [COMPLETED]
- Changes committed with session {session_id}
- Next: Install typst for PDF output, or review typst source / markdown report
Note: Typst source is always generated in Phase 4 regardless of typst CLI availability.
Phase 5 (PDF compilation) is optional -- if typst CLI is not installed or compilation fails,
the task still completes successfully. The .typ source file remains the primary artifact.
Expected partial return (core phase failure):
Founder implementation partially completed for task {N}:
- Phases {phases_completed}/{phases_total} executed
- Data gathered: {brief summary of progress}
- Resume phase: {resume_phase}
- Status remains [IMPLEMENTING]
- Next: Run /implement {N} to resume from phase {resume_phase}
Error Handling
Session ID Missing
Return immediately with failed status.
Plan Not Found
Return error with guidance to run /plan first.
Task Not Found
Return error with guidance to check task number.
Agent Errors
Pass through the agent's error return verbatim.
Build/Calculation Errors
Return partial status with progress made.
Phase 5 Typst/PDF Errors
Phase 5 failures do NOT block task completion:
- Typst not installed: Task completes with markdown output only
- Compilation error: Keep .typ file for debugging, task completes
- PDF empty: Keep .typ file, task completes
Postflight should check metadata.typst_source_generated and metadata.pdf_compiled to determine what artifacts to report.
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