Agent Skillsmtarcure/claude-vibe-squad › dual-level-retrieval

dual-level-retrieval

GitHub

一种双级检索策略,先广泛粗搜定位候选位置,再精读关键文件获取证据,确保引用准确。适用于搜索结果粗糙或需精准验证声明的场景。

.agents/skills/dual-level-retrieval/SKILL.md mtarcure/claude-vibe-squad

触发场景

搜索结果过于宽泛无法支撑结论 需要精准引用代码定义或调用关系

安装

npx skills add mtarcure/claude-vibe-squad --skill dual-level-retrieval -g -y
更多选项

非标准路径

npx skills add https://github.com/mtarcure/claude-vibe-squad/tree/main/.agents/skills/dual-level-retrieval -g -y

不安装直接使用

npx skills use mtarcure/claude-vibe-squad@dual-level-retrieval

指定 Agent (Claude Code)

npx skills add mtarcure/claude-vibe-squad --skill dual-level-retrieval -a claude-code -g -y

安装 repo 全部 skill

npx skills add mtarcure/claude-vibe-squad --all -g -y

预览 repo 内 skill

npx skills add mtarcure/claude-vibe-squad --list

SKILL.md

Frontmatter
{
    "name": "dual-level-retrieval",
    "audience": "specialist",
    "description": "Use when search results are too coarse to support a claim—sweep broadly to identify every candidate location, then open the most decisive hits around their definitions and callers before quoting or citing them."
}

Dual-Level Retrieval

Retrieve twice — once coarsely over the whole corpus to locate candidates, once precisely inside them to read evidence — so breadth never comes at the cost of quoting accurately.

Steps

  1. Run the coarse pass over the entire corpus using cheap, recall-oriented queries: names, symbols, and identifiers rather than concepts. Optimize for missing nothing, and accept false positives.
  2. Record the coarse hit set as a list of locations before reading any of them. This list is the denominator for the fine pass.
  3. Rank the hit set by expected decisiveness, not by match count — the file that defines a behavior outranks the ten that mention it.
  4. Run the fine pass by reading the ranked candidates in full context, wide enough to capture the enclosing definition and its callers.
  5. Quote from the fine pass only. Any claim traced to a coarse-pass snippet is unverified, because a grep line is not a reading of the code.
  6. Feed fine-pass discoveries back into a new coarse query when they reveal vocabulary you did not know to search for, and iterate until a coarse pass adds no new decisive locations.
  7. Report both levels: how many locations the coarse pass found, how many the fine pass read, and what was deliberately left unread.

Acceptance

  • Coarse queries and their exact hit counts are recorded.
  • Every quoted line comes from a file read in the fine pass.
  • Candidate ranking is justified by decisiveness rather than frequency.
  • At least one feedback iteration is attempted, or its absence is justified.
  • The gap between locations found and locations read is stated explicitly.

版本历史

  • d5262e2 当前 2026-09-11 11:16

同 Skill 集合

.agents/skills/accessible-media-authoring/SKILL.md
.agents/skills/agent-prompt-engineering/SKILL.md
.agents/skills/agentic-safety-audit/SKILL.md
.agents/skills/audio-event-map-authoring/SKILL.md
.agents/skills/auto-scaffold/SKILL.md
.agents/skills/claim-verification/SKILL.md
.agents/skills/code-reachability-audit/SKILL.md
.agents/skills/code-review-loop/SKILL.md
.agents/skills/color-theory/SKILL.md
.agents/skills/conversation-design/SKILL.md
.agents/skills/copy-refinement/SKILL.md
.agents/skills/cross-file-relationship-synthesis/SKILL.md
.agents/skills/dependency-cycle-audit/SKILL.md
.agents/skills/dependency-health-triage/SKILL.md
.agents/skills/detection-as-code/SKILL.md
.agents/skills/diff-aware-semgrep-scan/SKILL.md
.agents/skills/differential-review/SKILL.md
.agents/skills/dimensional-analysis-check/SKILL.md
.agents/skills/figma-implement-design/SKILL.md
.agents/skills/forensic-timeline-authoring/SKILL.md
.agents/skills/game-design-fundamentals/SKILL.md
.agents/skills/game-mechanics-balancing/SKILL.md
.agents/skills/head-tail/SKILL.md
.agents/skills/incident-response-runbook/SKILL.md
.agents/skills/interactive-audio-design/SKILL.md
.agents/skills/interface-ambiguity-check/SKILL.md
.agents/skills/keyword-clustering/SKILL.md
.agents/skills/knowledge-base-integration/SKILL.md
.agents/skills/layered-analysis-loop/SKILL.md
.agents/skills/level-design-patterns/SKILL.md
.agents/skills/locale-adaptation/SKILL.md
.agents/skills/narrative-structure/SKILL.md
.agents/skills/platform-compliance/SKILL.md
.agents/skills/player-engagement-psychology/SKILL.md
.agents/skills/requirements-elicitation/SKILL.md
.agents/skills/rule6-rights-gate/SKILL.md
.agents/skills/rule8-truth-gate/SKILL.md
.agents/skills/sandbox-provision-discipline/SKILL.md
.agents/skills/scope-decomposition/SKILL.md
.agents/skills/scope-estimation/SKILL.md
.agents/skills/security-ownership-map/SKILL.md
.agents/skills/security-threat-model/SKILL.md
.agents/skills/semgrep-rule-author/SKILL.md
.agents/skills/skill-description-trigger-authoring/SKILL.md
.agents/skills/sound-design-principles/SKILL.md
.agents/skills/structured-data-authoring/SKILL.md
.agents/skills/supply-chain-audit/SKILL.md
.agents/skills/take-over-resume/SKILL.md
.agents/skills/technical-seo-audit/SKILL.md

元信息

文件数
0
版本
d5262e2
Hash
2f6190dd
收录时间
2026-09-11 11:16

首页 - Wiki
Copyright © 2011-2026 iteam. Current version is 2.155.2. UTC+08:00, 2026-09-12 15:31
浙ICP备14020137号-1 $访客地图$