Agent skill
maintain-project
Automated maintenance for project-specific content (CLAUDE.md skills list, plans/ cleanup, health reports). TRIGGER when: user wants to maintain project documentation ("maintain project", "clean up plans", "update CLAUDE.md skills", "project health check"). DO NOT TRIGGER when: user wants framework sync (use /update-framework), or just wants to read documentation.
Install this agent skill to your Project
npx add-skill https://github.com/majiayu000/claude-skill-registry/tree/main/skills/other/other/maintain-project
SKILL.md
Maintain Project - Automated Project Content Maintenance
自动维护项目特定内容,保持 AI 编码基础设施健康运行。
Overview
What it does:
- 同步 CLAUDE.md 技能列表(从
.claude/skills/扫描已安装技能) - 清理
.claude/plans/目录(归档已完成计划,删除废弃计划) - 生成项目健康报告(0-100 分数 + 可操作建议)
Why it's needed: 随着项目演进,AI 编码基础设施逐渐退化:
- CLAUDE.md 中的技能列表过时(新增技能未列出)
- .claude/plans/ 中积累大量已完成计划(占用空间)
- 文档与实际代码不一致(降低 AI 辅助效率)
This skill automates maintenance to keep project content fresh and accurate.
When to use:
- Monthly project maintenance routine
- After installing new skills (sync to CLAUDE.md)
- When .claude/plans/active/ becomes cluttered
- Before major releases (ensure documentation accuracy)
- After completing multiple issues (clean up plans)
Scope Separation: maintain-project vs update-framework
Critical distinction:
| Aspect | /maintain-project | /update-framework |
|---|---|---|
| Layer | 项目层(Project Layer) | 框架层(Framework Layer) |
| What | CLAUDE.md, docs/, plans/, ADRs | skills, rules, workflow, pillars |
| Why | 维护项目特定内容 | 同步框架组件 |
| When | 项目演进后维护 | 框架升级时同步 |
| Scope | 当前项目内容 | 跨项目框架 |
Example:
# 项目层维护(本技能)
/maintain-project # 更新 CLAUDE.md 技能列表
/maintain-project --component plans # 清理 plans/
# 框架层同步(update-framework)
/update-framework --from ~/dev/ai-dev # 同步框架组件
No overlap:
- maintain-project NEVER modifies
.claude/skills/,.claude/rules/,.prot/pillars/ - update-framework NEVER modifies
CLAUDE.md,docs/,.claude/plans/
Core Modules (5)
1. CLAUDE.md Maintenance (Phase 1: MVP ✅)
功能:自动同步技能列表
检测项:
- 新增技能未列出(需要添加到 CLAUDE.md)
- 已删除技能仍列出(需要从 CLAUDE.md 移除)
- 技能描述不一致(需要同步)
- 版本号不匹配(需要更新)
工作流:
1. 扫描 .claude/skills/ 获取已安装技能
2. 解析 CLAUDE.md 当前技能列表
3. 检测差异(新增/删除/版本不匹配)
4. 自动更新 CLAUDE.md 技能表格
5. 保持分类结构(Issue Lifecycle, Planning, Quality 等)
示例输出:
CLAUDE.md Skills Sync
=====================
✅ Scanned 20 installed skills
✅ Found 2 new skills not documented:
- maintain-project v1.0.0 (Project Maintenance)
- sync-docs v1.1.0 (Documentation)
✅ Found 1 removed skill still documented:
- old-skill (deprecated)
✅ Updated CLAUDE.md skills table
2. docs/ Content Maintenance (Phase 2: Future)
功能:验证文档内容准确性
检测项:
- 代码示例过时(与实际代码不匹配)
- 链接失效(404 或文件不存在)
- API 文档与实现不一致
- 配置说明过期
工作流:
1. 读取 DOCUMENTATION_MANUAL.md(文档标准)
2. 扫描 docs/ 下所有 Markdown 文件
3. 提取代码块并验证(对比实际源码)
4. 检测内部链接有效性
5. 生成过时内容报告
Note: Phase 2 implementation, not included in MVP.
3. plans/ Cleanup (Phase 1: MVP ✅)
功能:清理计划目录
归档已完成计划:
# 检测 issue 状态
for plan in .claude/plans/active/*.md; do
issue_num=$(grep -oP 'Issue #\K\d+' "$plan" | head -1)
# 检查 issue 是否已关闭
if gh issue view "$issue_num" --json state | grep -q "CLOSED"; then
mv "$plan" .claude/plans/archived/
fi
done
删除废弃计划:
# 超过 30 天的已关闭 issue 计划
for plan in .claude/plans/archived/*.md; do
issue_num=$(extract_issue_number "$plan")
closed_date=$(gh issue view "$issue_num" --json closedAt -q .closedAt)
if days_since "$closed_date" > 30; then
rm "$plan"
fi
done
保持整洁:
- 只保留当前活跃 issue 的计划在 active/
- 已完成的移到 archived/
- 长时间未活动的删除
示例输出:
Plans Cleanup
=============
✅ Found 5 completed plans in active/
✅ Moved 5 plans to archived/
✅ Deleted 2 plans (closed > 30 days ago)
Before: 12 files in active/
After: 5 files in active/ (optimal)
4. ADRs Maintenance (Phase 2: Future)
功能:维护 ADR 状态和索引
检测项:
- ADR 状态未更新(Proposed → Accepted → Superseded)
- 索引文件过时(README.md 未包含新 ADR)
- 编号跳跃(缺失 ADR 编号)
- 状态不一致(文件名 vs 内容)
工作流:
1. 扫描 docs/ADRs/ 获取所有 ADR
2. 检测编号连续性
3. 解析每个 ADR 的状态字段
4. 更新 docs/ADRs/README.md 索引
5. 生成状态报告
Note: Phase 2 implementation, not included in MVP.
5. Health Report (Phase 1: MVP ✅)
功能:生成项目健康评分
评分计算:
def calculate_health_score():
scores = {
"claude_md": check_claude_md_health(), # 0-100
"plans": check_plans_cleanliness(), # 0-100
"overall": 0
}
# 加权平均(Phase 1: 仅 2 个组件)
scores["overall"] = (scores["claude_md"] * 0.5 +
scores["plans"] * 0.5)
return scores
def check_claude_md_health():
installed = scan_skills_directory(".claude/skills/")
documented = parse_skills_section("CLAUDE.md")
# 计算匹配率
match_rate = len(installed & documented) / len(installed)
return int(match_rate * 100)
def check_plans_cleanliness():
active_plans = count_files(".claude/plans/active/")
# 检查有多少应该归档
should_archive = count_completed_plans()
# 清洁度 = (总数 - 应归档) / 总数
cleanliness = (active_plans - should_archive) / active_plans if active_plans > 0 else 1.0
return int(cleanliness * 100)
报告格式:
Project Health Report
=====================
Overall Health: 85/100 ✅
Component Health:
- CLAUDE.md: 90/100 ✅
- 2 new skills not documented
- Skills section 90% accurate
- plans/: 80/100 ⚠️
- 5 completed plans should be archived
- active/ directory: 12 files (optimal: 5)
Recommendations:
1. Update CLAUDE.md skills list (2 additions)
2. Archive 5 completed plans
3. Run /maintain-project to auto-fix
AI Execution Instructions
CRITICAL: Task creation and component execution
When executing /maintain-project, AI MUST follow this workflow:
Step 1: Create 5 Subtasks (MVP)
# Create all tasks at the start
tasks = {
"validate": TaskCreate(
subject="Validate project structure",
description="Verify .claude/, CLAUDE.md, plans/ exist",
activeForm="Validating project..."
),
"claude_md": TaskCreate(
subject="Sync CLAUDE.md skills list",
description="Scan .claude/skills/ and update CLAUDE.md",
activeForm="Syncing CLAUDE.md..."
),
"plans": TaskCreate(
subject="Clean up plans/ directory",
description="Archive completed, delete abandoned",
activeForm="Cleaning plans/..."
),
"health": TaskCreate(
subject="Generate health report",
description="Calculate scores and recommendations",
activeForm="Calculating health..."
),
"summary": TaskCreate(
subject="Report comprehensive summary",
description="Display results and next steps",
activeForm="Generating summary..."
)
}
Step 2: Execute with Status Updates
# Validation
TaskUpdate(tasks["validate"], "in_progress")
validate_project_structure()
TaskUpdate(tasks["validate"], "completed")
# CLAUDE.md sync (unless --component specified)
if component in [None, "claude-md"]:
TaskUpdate(tasks["claude_md"], "in_progress")
sync_claude_md_skills()
TaskUpdate(tasks["claude_md"], "completed")
# Plans cleanup (unless --component specified)
if component in [None, "plans"]:
TaskUpdate(tasks["plans"], "in_progress")
cleanup_plans_directory()
TaskUpdate(tasks["plans"], "completed")
# Health report (always run unless --report-only)
TaskUpdate(tasks["health"], "in_progress")
health = generate_health_report()
TaskUpdate(tasks["health"], "completed")
# Summary
TaskUpdate(tasks["summary"], "in_progress")
print_summary(health)
TaskUpdate(tasks["summary"], "completed")
Step 3: Component-Specific Execution
CLAUDE.md sync implementation:
def sync_claude_md_skills():
# 1. 扫描已安装技能
installed_skills = []
for skill_dir in glob(".claude/skills/*/"):
skill_md = f"{skill_dir}/SKILL.md"
if exists(skill_md):
# 解析 YAML frontmatter
metadata = parse_yaml_frontmatter(skill_md)
installed_skills.append({
"name": metadata["name"],
"version": metadata.get("version", "unknown"),
"description": metadata.get("description", "").split("\n")[0]
})
# 2. 解析 CLAUDE.md 当前技能列表
current_skills = parse_skills_table_from_claude_md()
# 3. 检测差异
installed_names = {s["name"] for s in installed_skills}
current_names = {s["name"] for s in current_skills}
added = installed_names - current_names
removed = current_names - installed_names
# 4. 更新 CLAUDE.md
if added or removed:
update_skills_section("CLAUDE.md", installed_skills)
print(f"✅ Updated CLAUDE.md: +{len(added)}, -{len(removed)}")
else:
print("✅ CLAUDE.md skills list is up to date")
Plans cleanup implementation:
def cleanup_plans_directory():
active_dir = ".claude/plans/active/"
archived_dir = ".claude/plans/archived/"
# 确保 archived/ 目录存在
os.makedirs(archived_dir, exist_ok=True)
# 1. 归档已完成计划
archived_count = 0
for plan_file in glob(f"{active_dir}/*.md"):
# 提取 issue 编号
content = read_file(plan_file)
match = re.search(r'Issue #(\d+)', content)
if not match:
continue
issue_num = match.group(1)
# 检查 issue 状态
result = run_command(f"gh issue view {issue_num} --json state -q .state")
if result.strip() == "CLOSED":
# 移动到 archived/
shutil.move(plan_file, archived_dir)
archived_count += 1
# 2. 删除超过 30 天的归档计划(可选)
deleted_count = 0
for plan_file in glob(f"{archived_dir}/*.md"):
content = read_file(plan_file)
match = re.search(r'Issue #(\d+)', content)
if not match:
continue
issue_num = match.group(1)
# 获取关闭日期
result = run_command(f"gh issue view {issue_num} --json closedAt -q .closedAt")
if result.strip():
closed_date = datetime.fromisoformat(result.strip().rstrip('Z'))
days_ago = (datetime.now() - closed_date).days
if days_ago > 30:
os.remove(plan_file)
deleted_count += 1
print(f"✅ Plans cleanup: {archived_count} archived, {deleted_count} deleted")
Health report implementation:
def generate_health_report():
# CLAUDE.md 健康评分
installed = scan_skills_directory(".claude/skills/")
documented = parse_skills_section("CLAUDE.md")
claude_md_match_rate = len(installed & documented) / len(installed) if installed else 1.0
claude_md_score = int(claude_md_match_rate * 100)
# Plans 清洁度评分
active_plans = count_files(".claude/plans/active/")
should_archive = count_completed_plans()
plans_cleanliness = (active_plans - should_archive) / active_plans if active_plans > 0 else 1.0
plans_score = int(plans_cleanliness * 100)
# 整体评分(加权平均)
overall_score = int((claude_md_score * 0.5 + plans_score * 0.5))
return {
"overall": overall_score,
"claude_md": {
"score": claude_md_score,
"issues": list(installed - documented),
},
"plans": {
"score": plans_score,
"should_archive": should_archive,
"active_count": active_plans
}
}
Workflow Steps (9 Tasks)
Phase 1: MVP Implementation
-
✅ Create skill structure (Task 1)
.claude/skills/maintain-project/SKILL.md.claude/skills/maintain-project/LICENSE.txt- Follow ADR-001 official patterns
-
✅ Implement CLAUDE.md skills sync (Task 2)
- Scan
.claude/skills/directory - Parse CLAUDE.md skills table
- Detect differences (new/removed/updated)
- Update CLAUDE.md automatically
- Scan
-
✅ Implement plans/ cleanup (Task 3)
- Archive completed plans (check GitHub issue status)
- Delete abandoned plans (closed > 30 days)
- Keep active/ directory clean
-
✅ Implement health report (Task 4)
- Calculate component scores (0-100)
- Generate overall health score
- Provide actionable recommendations
-
✅ Add CLI interface (Task 5)
--dry-run- Preview mode--component <name>- Run specific component--report-only- Health report only--verbose- Detailed logging
-
✅ Write documentation (Task 6)
- Complete SKILL.md (this file)
- Usage examples
- Error handling guide
- Integration with /update-framework
-
⏳ Create evaluation tests (Task 7)
- Test CLAUDE.md sync
- Test plans cleanup
- Test health report
- Test dry-run mode
-
⏳ Update CLAUDE.md (Task 8)
- Add maintain-project to skills table
- Add use case example
- Document integration
-
⏳ Integration testing (Task 9)
- Test in real project (buffer)
- Verify no framework files modified
- Verify accurate health scores
- Document results
Arguments
/maintain-project # 完整维护(CLAUDE.md + plans/ + health report)
/maintain-project --dry-run # 预览模式(不做修改,仅显示将执行的操作)
/maintain-project --component claude-md # 仅同步 CLAUDE.md
/maintain-project --component plans # 仅清理 plans/
/maintain-project --report-only # 仅生成健康报告(不做修改)
/maintain-project --verbose # 显示详细日志
参数说明:
| Argument | Description | Example |
|---|---|---|
--dry-run |
预览模式,不做任何修改 | --dry-run |
--component <name> |
仅运行指定组件(claude-md, plans) | --component plans |
--report-only |
仅生成健康报告 | --report-only |
--verbose |
显示详细执行日志 | --verbose |
组件名称:
claude-md- CLAUDE.md 技能列表同步plans- plans/ 目录清理
Phase 2 (未来):
docs- docs/ 内容验证adrs- ADRs 维护
Usage Examples
Example 1: Monthly Maintenance (Recommended)
User says:
"maintain project"
What happens:
🚀 Starting project maintenance...
Task Progress:
✅ Task 1/5: Validate project structure
✅ Task 2/5: Sync CLAUDE.md skills list
- Found 2 new skills: maintain-project, sync-docs
- Updated CLAUDE.md skills table
✅ Task 3/5: Clean up plans/ directory
- Archived 5 completed plans
- Deleted 2 abandoned plans
✅ Task 4/5: Generate health report
✅ Task 5/5: Report summary
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Project Health Report
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Overall Health: 95/100 ✅
Component Health:
- CLAUDE.md: 100/100 ✅
- All skills documented
- Versions up to date
- plans/: 90/100 ✅
- active/ directory: 5 files (optimal)
- 0 plans need archiving
Recommendations:
✅ No issues found - project is healthy!
Summary:
- CLAUDE.md: 2 skills added
- plans/: 5 archived, 2 deleted
- Health: 82/100 → 95/100 (+13 points)
Time: 8 seconds
Example 2: Preview Changes (Dry Run)
User says:
"show me what maintain-project would do"
What happens:
/maintain-project --dry-run
🔍 Preview Mode (no changes will be made)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
CLAUDE.md Sync Preview
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Would add 2 skills:
- maintain-project v1.0.0 (Project Maintenance)
- sync-docs v1.1.0 (Documentation)
Would remove 1 skill:
- old-skill (deprecated)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Plans Cleanup Preview
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Would archive 5 plans:
- issue-123-plan.md (closed 2 days ago)
- issue-124-plan.md (closed 5 days ago)
- issue-125-plan.md (closed 7 days ago)
- issue-126-plan.md (closed 10 days ago)
- issue-127-plan.md (closed 12 days ago)
Would delete 2 abandoned plans:
- issue-100-plan.md (closed 45 days ago)
- issue-101-plan.md (closed 60 days ago)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Health Report
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Current Health: 82/100 ⚠️
After Fix: 95/100 ✅ (+13 points)
Run without --dry-run to apply changes.
Example 3: CLAUDE.md Sync Only
User says:
"update CLAUDE.md skills list"
What happens:
/maintain-project --component claude-md
✅ Scanning .claude/skills/ directory...
✅ Found 20 installed skills
✅ Parsing CLAUDE.md current skills table...
✅ Found 18 documented skills
Differences:
+ maintain-project v1.0.0 (new)
+ sync-docs v1.1.0 (new)
✅ Updated CLAUDE.md skills table
Health: CLAUDE.md 90/100 → 100/100 ✅
Time: 3 seconds
Example 4: Plans Cleanup Only
User says:
"clean up plans directory"
What happens:
/maintain-project --component plans
✅ Scanning .claude/plans/active/...
✅ Found 12 active plans
Checking GitHub issue status...
- Issue #123: CLOSED → archive
- Issue #124: CLOSED → archive
- Issue #125: OPEN → keep
- Issue #126: CLOSED → archive
- ...
✅ Archived 5 completed plans
✅ Deleted 2 abandoned plans (closed > 30 days)
Before: 12 files in active/
After: 5 files in active/ ✅
Health: plans/ 70/100 → 90/100 (+20 points)
Time: 5 seconds
Example 5: Health Report Only
User says:
"check project health"
What happens:
/maintain-project --report-only
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Project Health Report
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
Overall Health: 82/100 ⚠️
Component Health:
- CLAUDE.md: 90/100 ✅
- 2 new skills not documented
- Skills section 90% accurate
- plans/: 75/100 ⚠️
- 5 completed plans should be archived
- 2 abandoned plans detected
- active/ directory: 12 files (optimal: 5)
Recommendations:
1. Run /maintain-project --component claude-md (adds 2 skills)
2. Run /maintain-project --component plans (archives 5, deletes 2)
3. Or run /maintain-project to fix all issues
Estimated improvement: 82/100 → 95/100 (+13 points)
Time: 2 seconds
Error Handling
Missing Project Structure
❌ Error: Invalid project structure
Missing required components:
- .claude/ directory not found
- CLAUDE.md not found
Please ensure this is a valid ai-dev project.
No GitHub CLI Access
❌ Error: GitHub CLI not available
Plans cleanup requires GitHub CLI (gh) to check issue status.
Install: brew install gh
Login: gh auth login
Or skip plans cleanup: /maintain-project --component claude-md
No Changes Needed
✅ Project is already healthy!
Current Health: 100/100 ✅
Component Status:
- CLAUDE.md: 100/100 ✅ (all skills documented)
- plans/: 100/100 ✅ (active/ directory clean)
No maintenance actions needed.
Dry Run Mode
🔍 Dry Run Mode (no changes will be made)
Preview:
- Would add 2 skills to CLAUDE.md
- Would archive 5 plans
- Would delete 2 abandoned plans
Run without --dry-run to apply changes.
Integration with /update-framework
Clear separation of concerns:
# 步骤 1: 更新框架组件(框架层)
/update-framework --from ~/dev/ai-dev
# → 同步 skills, rules, workflow, pillars
# 步骤 2: 维护项目内容(项目层)
/maintain-project
# → 更新 CLAUDE.md, 清理 plans/, 生成健康报告
Recommended workflow:
# Monthly routine
cd ~/projects/my-app
# 1. 拉取框架更新
/update-framework --from ~/dev/ai-dev --dry-run # 预览
/update-framework --from ~/dev/ai-dev # 执行
# 2. 维护项目内容
/maintain-project --report-only # 检查健康
/maintain-project # 修复问题
# 3. 提交更新
git add CLAUDE.md .claude/plans/
git commit -m "chore: monthly framework update + project maintenance"
No overlap:
/update-framework修改:.claude/skills/,.claude/rules/,.prot/pillars/,.claude/workflow//maintain-project修改:CLAUDE.md,.claude/plans/active/,.claude/plans/archived/
Performance
Expected execution time:
| Operation | Time | Notes |
|---|---|---|
| Full maintenance | 5-10s | CLAUDE.md + plans + health |
| CLAUDE.md only | 2-3s | Scan + parse + update |
| Plans only | 3-5s | Check GitHub issues |
| Health report only | 1-2s | Calculate scores |
| Dry run | 3-5s | No modifications |
Scalability:
| Project Size | Skills Count | Plans Count | Time |
|---|---|---|---|
| Small | 10 | 5 | 3s |
| Medium | 20 | 15 | 6s |
| Large | 30 | 30 | 10s |
Optimization:
- Parallel GitHub API calls for plan status checks
- Cache installed skills list during execution
- Skip unchanged files in CLAUDE.md sync
Best Practices
- 定期维护(推荐每月):
# Monthly routine
/maintain-project --report-only # 先检查健康
/maintain-project # 修复问题
- 新增技能后立即同步:
# 安装新技能
/update-skills --from ~/dev/ai-dev --skills maintain-project
# 同步到 CLAUDE.md
/maintain-project --component claude-md
- 完成多个 issue 后清理:
# 关闭多个 issue 后
/maintain-project --component plans
- 大型变更前预览:
/maintain-project --dry-run
# 确认无误后执行
/maintain-project
- 与框架更新配合:
# 框架更新后立即维护
/update-framework --from ~/dev/ai-dev
/maintain-project
Task Management
After each maintenance step, update progress:
Project validated → Update Task #1
CLAUDE.md synced → Update Task #2
Plans cleaned → Update Task #3
Health calculated → Update Task #4
Summary reported → Update Task #5
Provides real-time visibility of maintenance progress.
Final Verification
Before declaring maintenance complete, verify:
- [ ] All 5 tasks completed
- [ ] CLAUDE.md skills list accurate
- [ ] plans/active/ contains only active plans
- [ ] plans/archived/ contains completed plans
- [ ] Health score calculated
- [ ] No framework files modified (.claude/skills/, .claude/rules/)
- [ ] Recommendations provided
Missing items indicate incomplete maintenance.
Related Skills
Project Layer (complementary):
- /check-docs - Validate docs/ structure
- /init-docs - Create initial docs/ structure
- /adr - Manage Architecture Decision Records
Framework Layer (separate):
- /update-framework - Sync framework components (meta-skill)
- /update-skills - Sync skills only
- /update-rules - Sync rules only
- /update-workflow - Sync workflow docs only
- /update-pillars - Sync pillars only
Workflow Integration:
- /work-issue - Uses maintain-project after issue completion
- /finish-issue - May trigger maintain-project for cleanup
Version: 1.0.0 Last Updated: 2026-03-15 Changelog:
- v1.0.0 (2026-03-15): Initial release - project maintenance and cleanup skill Phase: 1 (MVP - CLAUDE.md, plans/, health) Pattern: Workflow Skill (project maintenance automation) Compliance: ADR-001 ✅
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
agent-ops-spec
Manage specification documents in .agent/specs/. Use when user provides requirements, acceptance criteria, or feature descriptions that need to be tracked and validated against implementation.
agent-ops-state
Maintain .agent state files. Use at session start, after meaningful steps, and before concluding: read/update constitution/memory/focus/issues/baseline consistently.
agent-ops-spec
Manage specification documents in .agent/specs/. Use when user provides requirements, acceptance criteria, or feature descriptions that need to be tracked and validated against implementation.
agent-ops-testing
Test strategy, execution, and coverage analysis. Use when designing tests, running test suites, or analyzing test results beyond baseline checks.
agent-ops-testing
Test strategy, execution, and coverage analysis. Use when designing tests, running test suites, or analyzing test results beyond baseline checks.
agent-ops-state
Maintain .agent state files. Use at session start, after meaningful steps, and before concluding: read/update constitution/memory/focus/issues/baseline consistently.
Didn't find tool you were looking for?