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

62%

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

The body presents a well-sequenced iterative workflow with explicit checkpoints (workflow_clarity scores 3), but spends tokens on conceptual framing, ships pseudocode rather than executable code, and keeps all detail inline with no progressive disclosure into reference files. The other three dimensions sit at 2.

Suggestions

Tighten or remove the conceptual '問題'/'標準方法失敗' framing and the decorative ASCII box — lead with the pattern itself rather than explaining the subagent context problem Claude already knows.

Make the code executable by defining or stubbing the helper functions (retrieveFiles, scoreRelevance, extractPatterns, hasCriticalGaps, mergeContext), or explicitly label the snippets as illustrative pseudocode.

Move the two worked examples and/or the full phase code into a separate reference file (e.g., EXAMPLES.md) linked from a concise overview to enable true one-level-deep progressive disclosure.

Dimension	Reasoning	Score
Conciseness	The phase code is lean, but the '問題' and '標準方法失敗' sections plus the decorative ASCII box explain concepts Claude already knows and add tokens that could be trimmed, matching 'mostly efficient but includes some unnecessary explanation'.	2 / 3
Actionability	Each phase has concrete code structure, but the functions call undefined helpers (retrieveFiles, scoreRelevance, extractPatterns, hasCriticalGaps, mergeContext), making it structured pseudocode rather than copy-paste executable, matching the score-2 anchor.	2 / 3
Workflow Clarity	A clearly sequenced 4-phase cycle (DISPATCH→EVALUATE→REFINE→LOOP) with an explicit EVALUATE scoring checkpoint, a gap-driven feedback loop, and a stop condition (≥3 high-relevance files and no critical gaps) — explicit checkpoints with a feedback loop as the score-3 anchor describes.	3 / 3
Progressive Disclosure	Sections are well organized, but it is a single ~200-line monolithic file with no bundle files or one-level-deep references; the worked examples and full phase code are all inline, matching 'content that should be separate is inline'.	2 / 3
	Total	9 / 12 Passed

Description

50%

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 clearly states the pattern's purpose and domain but stops short of the strongest level on every dimension: it names one abstract action rather than several concrete ones, lacks natural trigger phrasing, and omits any explicit 'when to use' guidance. All four dimensions land at 2.

Suggestions

Add an explicit 'Use when...' trigger clause (e.g., 'Use when a subagent needs relevant files but you don't yet know which files or terms are relevant') to raise completeness and trigger_term_quality.

Include natural user-facing keywords alongside the technical phrasing (e.g., 'finding relevant files for an agent task', 'retrieving context for subagents') so users would naturally say them.

Name the concrete actions the pattern performs (dispatch broad queries, score file relevance, identify gaps, refine) instead of the single abstract action 'refining context retrieval'.

Dimension	Reasoning	Score
Specificity	Names the domain and a single action — 'progressively refining context retrieval' — but does not list multiple concrete actions like the score-3 anchor; it stays at the 'names domain and some actions, not comprehensive' level.	2 / 3
Completeness	It states what the skill does (a pattern for refining retrieval) but gives no 'when to use it' guidance, so per the judging guideline a missing 'Use when...' clause caps completeness at 2.	2 / 3
Trigger Term Quality	Relevant keywords ('context retrieval', 'subagent') are present but lean technical ('subagent context problem'); it lacks natural user phrasing and a 'Use when' trigger, matching 'some relevant keywords but missing common variations'.	2 / 3
Distinctiveness Conflict Risk	The 'subagent context problem' niche is somewhat specific, but without explicit distinct triggers it could still overlap with other retrieval/context skills, fitting 'somewhat specific but could still overlap'.	2 / 3
	Total	8 / 12 Passed

Validation

100%

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 — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository: affaan-m/everything-claude-code
Commit: 4130457

Reviewed: 4 days 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.