Token导航 LogoToken导航TokenDH.com
研究检索只读github未标认证来源可访问许可证需确认审计提醒

vulture-sweep秃鹫扫掠

Agent Skill

vulture-sweep 用于查找、检索和筛选相关信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

总安装

1,388

周安装

59

GitHub Stars

4

下载量

486
CodexClaudeCursorGemini CLI

安装说明

本站只整理中文说明和来源信息,不托管安装包,也不代用户安装。

GitHub

来源数

2

许可证

unknown

最后核验

2026-05-01

来源状态

来源可访问

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:vulture-sweep(秃鹫扫掠)
来源仓库:https://github.com/autumnsgrove/groveengine
仓库路径:skills/vulture-sweep
安装命令:
npx skills add https://github.com/autumnsgrove/groveengine --skill vulture-sweep
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。该命令会通过 npx skills 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

skills.shnpx skills
npx skills add https://github.com/autumnsgrove/groveengine --skill vulture-sweep

简介

用于系统性扫描代码库中的潜在问题与模式异常。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

  • 适合发现重复代码、复杂函数或不符合规范的写法。
  • 通过 GitHub 安装,支持自定义规则集扩展。
  • 扫描结果应按优先级排序,聚焦高风险项先行处理。
  • vulture-sweep 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

SKILL.md

Vulture Sweep 🦅

The vulture circles patiently above, seeing what others miss. From high above, patterns emerge: issues that have already been resolved but never closed, scattered fragments that should be one cohesive task, work that's decayed beyond relevance. The vulture doesn't judge. It performs a necessary service that others avoid—nature's cleanup crew. When the vulture descends, it verifies before consuming. When it rises again, the backlog is healthy.

When to Activate

  • Issue board has accumulated cruft over time
  • User says "clean up my issues" or "sweep the backlog"
  • User calls /vulture-sweep or mentions vulture/cleanup
  • Suspecting issues are done but never closed
  • Finding scattered related issues that should be consolidated
  • Quarterly or monthly backlog hygiene
  • After a major release (catch implemented-but-not-closed issues)

IMPORTANT: This animal NEVER closes issues without codebase verification. It only closes what has been proven complete.

Pair with: bee-collect when finding work that needs new issues, badger-triage for organizing what remains


The Sweep

CIRCLE → SPOT → DESCEND → CLEAN → REPORT
   ↓        ↓        ↓         ↓        ↓
 Survey   Identify  Verify   Close/   Summary
 Board    Stale     in Code  Consol.  & Stats

Phase 1: CIRCLE

*The vulture rises on thermal currents, patient eyes scanning the landscape below...*

Survey the entire issue board:

Gather all open issues:

# Get all open issues with metadata
gh issue list --repo AutumnsGrove/Lattice --state open --limit 200 --json number,title,labels,createdAt,updatedAt,body

Build the mental map:

  • Group by age (>90 days old = potential decay)
  • Group by labels (component clusters)
  • Group by keywords in titles (similar work patterns)
  • Note issues with "implemented", "done", "completed" in comments

Age thresholds:

AgeStatus
< 30 daysFresh
30-90 daysAging
90-180 daysStale
> 180 daysDecaying

Output: Complete map of the issue board with age and pattern analysis


Phase 2: SPOT

*Sharp eyes catch movement. Something dead lies in the grass...*

Identify candidates for cleanup:

Category 1: Likely Implemented (Dead) Look for signals that work may be done:

  • Old issues with recent related commits
  • Issues mentioning features that now exist
  • Bug reports for bugs that may be fixed
  • Enhancement requests that match current behavior

Category 2: Stale/Outdated (Decaying) Look for signals of irrelevance:

  • References to deprecated technology
  • Issues about removed features
  • Problems that no longer apply to current architecture
  • Outdated acceptance criteria

Category 3: Fragmented (Scattered Bones) Look for consolidation opportunities:

  • Multiple issues about the same component
  • Overlapping acceptance criteria
  • Issues that are subtasks of a larger effort
  • Duplicates with slightly different wording

Candidate List Format:

## Candidates Spotted

### Likely Implemented

| #    | Title                | Age  | Signal                    |
| ---- | -------------------- | ---- | ------------------------- |
| #234 | Add dark mode toggle | 120d | Feature exists in Foliage |
| #267 | Fix login redirect   | 85d  | Recent auth commits       |

### Potentially Stale

| #    | Title                 | Age  | Signal         |
| ---- | --------------------- | ---- | -------------- |
| #189 | Update webpack config | 200d | Now using Vite |

### Consolidation Candidates

| Issues           | Theme               | Recommendation            |
| ---------------- | ------------------- | ------------------------- |
| #301, #305, #312 | Accessibility fixes | Combine into "A11y audit" |

Output: Categorized candidates ready for verification


Phase 3: DESCEND

*The vulture folds its wings and drops, examining closely what it found from above...*

Verify each candidate in the codebase:

For "Likely Implemented" candidates:

# Grove Find — verify implementation exists
gf --agent search "darkMode"          # Does the feature exist in code?
gf --agent search "ThemeToggle"       # Find the component
gf --agent func "toggleDarkMode"      # Find the function

# Check git history for evidence
git log --oneline --all --grep="dark mode" -10
git log --oneline --all -- "src/lib/stores/theme*"

Verification Checklist:

  • Core functionality exists in code
  • No TODO comments indicating incomplete work
  • Tests pass (if applicable)
  • Feature accessible in the UI (if applicable)

Verification Outcomes:

OutcomeEvidence RequiredAction
Fully DoneCode exists, tests pass, feature worksClose with detailed comment
Partially DoneSome code exists, more neededKeep open, update description
Not StartedNo evidence foundKeep open
ObsoleteFeature no longer relevantClose as "won't do" with explanation

For "Fragmented" candidates:

  • Read all related issues
  • Identify the unifying theme
  • Draft consolidated issue description
  • List which issues would be closed

Output: Verified candidates with evidence documented


Phase 4: CLEAN

*The vulture consumes what is dead, leaving the ecosystem healthier...*

Execute the cleanup:

Grove Tools for bulk cleanup:

# Single issue close
gw gh issue close --write NUMBER --comment "Verified complete — evidence in src/lib/..."

# Bulk operations — close multiple verified issues at once
gw gh issue batch --write --close 234,267,289 --comment "Vulture sweep: verified implemented"

Closing Implemented Issues (with detailed comments):

gh issue close NUMBER --comment "$(cat <<'EOF'
🦅 **Vulture Sweep: Verified Complete**

This issue has been implemented and verified in the codebase:

**Evidence Found:**
- Feature implemented in `src/lib/stores/theme.ts`
- Dark mode toggle exists in `src/lib/components/ThemeToggle.svelte`
- Tests passing in `tests/theme.test.ts`
- Verified working in production UI

**Commits:**
- abc1234 feat(foliage): add dark mode support
- def5678 test(foliage): add theme toggle tests

Closing as complete. Thank you for the contribution!
EOF
)"

Closing Obsolete Issues:

gh issue close NUMBER --reason "not planned" --comment "$(cat <<'EOF'
🦅 **Vulture Sweep: No Longer Applicable**

This issue is being closed as it no longer applies to the current codebase:

**Reason:**
- Grove migrated from Webpack to Vite in January 2026
- Webpack configuration no longer exists
- The underlying problem this addressed is obsolete

If similar issues arise with the current Vite setup, please open a new issue.
EOF
)"

Creating Consolidated Issues:

gh issue create --title "Consolidated: Accessibility improvements" --body "$(cat <<'EOF'
## Summary

This issue consolidates several related accessibility tasks into a single trackable effort.

## Consolidated From

- #301 — Add aria labels to navigation
- #305 — Fix focus ring visibility
- #312 — Improve screen reader announcements

## Acceptance Criteria

- [ ] All navigation elements have appropriate aria labels
- [ ] Focus rings are visible in all themes
- [ ] Screen reader announcements work for dynamic content
- [ ] WCAG AA compliance verified

## Context

These issues were identified during vulture sweep as fragments of a larger accessibility effort. Consolidating for clearer tracking.
EOF
)" --label "accessibility,enhancement"

# Close the original issues with reference
gh issue close 301 --comment "🦅 Consolidated into #NEW_NUMBER"
gh issue close 305 --comment "🦅 Consolidated into #NEW_NUMBER"
gh issue close 312 --comment "🦅 Consolidated into #NEW_NUMBER"

Output: Issues closed, consolidations created, backlog cleaned


Phase 5: REPORT

*The vulture rises again, circling once more to survey the cleaner landscape...*

Report the sweep results:

🦅 VULTURE SWEEP COMPLETE

## Summary

| Action                            | Count |
| --------------------------------- | ----- |
| Issues Closed (Implemented)       | 7     |
| Issues Closed (Obsolete)          | 3     |
| Consolidations Created            | 2     |
| Issues Kept (Verified Incomplete) | 5     |

## Issues Closed as Implemented

| #    | Title                | Evidence                  |
| ---- | -------------------- | ------------------------- |
| #234 | Add dark mode toggle | Feature in theme.ts       |
| #267 | Fix login redirect   | Verified in auth flow     |
| #289 | Add loading states   | Skeleton components exist |
| ...  | ...                  | ...                       |

## Issues Closed as Obsolete

| #    | Title                 | Reason                   |
| ---- | --------------------- | ------------------------ |
| #189 | Update webpack config | Migrated to Vite         |
| #201 | Fix IE11 support      | IE11 no longer supported |
| #215 | Update Node 14 deps   | Now on Node 20           |

## Consolidations

| New Issue | Absorbed         | Theme             |
| --------- | ---------------- | ----------------- |
| #534      | #301, #305, #312 | Accessibility     |
| #535      | #298, #303       | Mobile navigation |

## Remaining Cleanup Opportunities

These issues may need attention but require human decision:

- #245 "Improve performance" — too vague to verify
- #278 "Consider alternative auth" — needs design decision
- #291 "Maybe add feature X" — unclear if still wanted

## Backlog Health

| Metric            | Before  | After   |
| ----------------- | ------- | ------- |
| Open Issues       | 87      | 72      |
| Average Age       | 95 days | 67 days |
| Issues > 180 days | 12      | 3       |

_The carrion is cleared. The forest breathes easier._

Vulture Rules

Patience

Circle first. Never swoop without surveying. The full picture matters.

Verification

Always verify before closing. The vulture's reputation depends on accuracy. Never close an issue without codebase evidence that it's truly complete or obsolete.

Respect

Close with detailed comments. The original reporter deserves to know *why* their issue is being closed and *what* was found.

Consolidation Over Fragmentation

When you find scattered related issues, consolidate them. One clear issue beats five overlapping ones.

Communication

Use sweep metaphors:

  • "Circling the board..." (surveying)
  • "Spotting candidates..." (identifying)
  • "Descending to verify..." (checking codebase)
  • "Cleaning up..." (closing/consolidating)
  • "Rising to report..." (summary)

Anti-Patterns

The vulture does NOT:

  • Close issues without codebase verification
  • Assume something is done because it's old
  • Touch issues marked "in progress" or assigned
  • Delete issues (only closes with clear reasoning)
  • Close issues that need human decision (marks for review instead)
  • Guess at implementation status
  • Skip the detailed closure comment

Example Sweep

User: "/vulture-sweep — the backlog is getting crusty"

Vulture flow:

  1. 🦅 CIRCLE — "Circling the board... 87 open issues found. 12 over 180 days old. 23 between 90-180 days. Patterns emerging in auth, UI, and infrastructure clusters."
  2. 🦅 SPOT — "Spotted 15 candidates:

- 7 likely implemented (dark mode, auth fixes, loading states) - 3 potentially obsolete (webpack, IE11, Node 14) - 5 fragmentation opportunities (a11y cluster, mobile nav)"

  1. 🦅 DESCEND — "Descending to verify...

- #234 dark mode: VERIFIED (found in theme.ts, toggle component exists) - #267 login redirect: VERIFIED (auth flow handles it) - #245 'improve performance': CANNOT VERIFY (too vague, keeping open)"

  1. 🦅 CLEAN — "Closing 10 verified issues with detailed comments. Creating 2 consolidations. Marking 5 for human review."
  2. 🦅 REPORT — "Sweep complete: 15 issues resolved, 2 consolidations created, backlog reduced from 87 to 72. Average age improved from 95 to 67 days."

Quick Decision Guide

SituationApproach
Issue clearly doneVerify in code, close with evidence
Issue partially doneKeep open, update description with findings
Issue obsoleteClose as "not planned" with explanation
Multiple related issuesConsolidate into one, close originals with reference
Issue vague/unclearKeep open, flag for human review
Issue has assigneeSkip (someone's working on it)
Recent issue (< 30 days)Skip unless clearly implemented

Integration with Other Skills

Before Sweeping:

  • bloodhound-scout — If you need to understand the codebase first

During Sweeping:

  • bee-collect — If you find new work that needs issues
  • badger-triage — If remaining issues need organizing

After Sweeping:

  • owl-archive — To document any patterns discovered
  • gathering-planning — If cleanup revealed planning needs

*Nature's cleanup crew. Patient, thorough, necessary.* 🦅

适合场景

01

用户想查找某类 Agent Skill 时

02

需要根据任务场景推荐可安装能力包时

03

需要对比不同来源的安装命令和来源信息时

能力概览

能力 1

按任务关键词查找相关 Skills

能力 2

展示可复制的安装命令

能力 3

保留来源站点、仓库和原始说明,方便继续核验

能力 4

展示第三方安全扫描或审计结果

安装后应在对应宿主中按原始 README 的触发条件使用;具体调用方式请以来源页面和 README 为准。

平台分布

Codex

35.92%
按下载量换算175

Claude

28.91%
按下载量换算141

Cursor

17.14%
按下载量换算83

Gemini CLI

10.06%
按下载量换算49

安全审计

Gen Agent Trust Hub

通过

Socket

通过

Snyk

可疑

权限和风险

只读

该 Skill 主要提供规则、说明或参考内容,本身偏只读;真正读写文件、联网或执行命令仍取决于宿主 Agent 的任务。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在 warning/failed 结果,不能写成本站确认安全。当前只有一个来源,正式发布前建议补源仓库或其他目录站核验。

来源信息

继续浏览同类 Skills