iterative-retrieval

Pattern for progressively refining context retrieval to solve the subagent context problem

Quality

—

Does it follow best practices?

Impact

—

No eval scenarios have been run

Securityby

Passed

No known issues

Quality

Content

85%

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

The content delivers a well-structured, actionable four-phase workflow with executable code, explicit stopping conditions, and clean organization. Its only weakness is mild verbosity from restating familiar concepts and a redundant integration section.

Suggestions

Tighten "The Problem" section: drop the enumeration of what subagents don't know (Claude already knows this) and keep only the failure modes that motivate the iterative approach.

Trim the "Integration with Agents" section, which restates the same four phases already detailed above, or fold it into a single one-line prompt snippet.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient with executable code, but sections like "The Problem" restate concepts Claude already knows and "Integration with Agents" repeats the loop steps, so some tokens could be trimmed to reach the lean anchor.	2 / 3
Actionability	All four phases include concrete, executable JavaScript with real functions (retrieveFiles, evaluateRelevance, refineQuery, iterativeRetrieve) plus explicit scoring criteria, making the guidance copy-paste ready.	3 / 3
Workflow Clarity	The DISPATCH→EVALUATE→REFINE→LOOP sequence is clearly ordered with an explicit sufficiency checkpoint (highRelevance.length >= 3 && !hasCriticalGaps) and a bounded max-3-cycles feedback loop for error recovery.	3 / 3
Progressive Disclosure	The single-file body is well organized into clearly signaled sections with no nested reference chains; with no bundle files, this matches the well-organized-sections anchor for a self-contained skill.	3 / 3
	Total	11 / 12 Passed

Description

35%

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

The description identifies a clear domain but is written in abstract, jargon-heavy language with no natural user triggers and no explicit "Use when" guidance. It is distinguishable but would benefit from concrete actions and trigger phrasing.

Suggestions

Add a "Use when..." clause with concrete user-facing triggers, e.g. "Use when spawning subagents that lack codebase context, or when retrieval returns too much or too little context."

Replace jargon ("subagent context problem") with natural terms users would actually say, such as "subagents missing context" or "retrieving the right files for a subagent".

Name concrete actions the skill performs, e.g. "iteratively search, score, and refine retrieved files" instead of the abstract "progressively refining context retrieval".

Dimension	Reasoning	Score
Specificity	The phrase "progressively refining context retrieval" names the domain and an implied action, but lists no concrete operations (no verbs like extract, merge, or search), landing between the vague and comprehensive anchors rather than at either.	2 / 3
Completeness	It states what the skill is ("Pattern for progressively refining context retrieval") but has no "Use when..." clause, so the when is only implied; per the guideline this caps completeness at 2.	2 / 3
Trigger Term Quality	Terms like "progressively refining context retrieval" and "subagent context problem" are technical jargon a user would almost never say naturally, matching the lowest anchor of technical jargon over natural keywords.	1 / 3
Distinctiveness Conflict Risk	The subagent-context niche is somewhat specific, but without explicit trigger terms it could still overlap with general RAG or retrieval skills, so it does not reach the clear-distinct-niche anchor.	2 / 3
	Total	7 / 12 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
metadata_version	'metadata.version' is missing	Warning

	Total	15 / 16 Passed

Repository: ysyecust/everything-claude-code
Commit: a9cbecb

Reviewed: 1 day ago

Table of Contents

Discovery Implementation Validation

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.