deep-research

Multi-agent research orchestration: split a research goal into parallel sub-goals, run each via headless `claude -p` subprocesses, aggregate results into a polished report file. Use for systematic web/document research, competitive or industry analysis, batch link/dataset processing, and long-form evidence synthesis. Triggers: "深度调研", "deep research", "wide research", "多 Agent 调研", "系统调研".

Quality

88%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Quality

Content

77%Weight 40%Scale 1-5

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

A thorough, highly actionable orchestration playbook with a well-sequenced workflow and strong validation loops. Its main weakness is token efficiency and structure: redundant tool-priority restatements, a basic-usage reference section that over-explains, and a single monolithic file where reference material could be split into bundled files.

Suggestions

State the tool-priority rule (skills → firecrawl → exa → WebFetch/WebSearch) once in the constraints section and reference it elsewhere instead of repeating the full chain 4–5 times.

Move the 'Claude Code 非交互模式参考' basic-usage and scheduling/parallel template blocks into a bundled reference file (e.g. references/claude_p_usage.md) and link to it from SKILL.md to improve progressive disclosure and reduce inline tokens.

Trim or remove the basic `claude -p "..."`, `--continue`, and `--resume` flag examples that Claude already knows, keeping only the skill-specific parameter patterns.

Dimension	Reasoning	Score
Conciseness	Mostly actionable procedural guidance rather than generic concept explanation, but the tool-priority chain ('skills → firecrawl → exa → WebFetch/WebSearch') is restated 4–5 times and the 'Claude Code 非交互模式参考' basic-usage section restates flags Claude largely already knows.	3 / 5
Actionability	Copy-paste-ready bash scripts with exact flags (`timeout 600 claude -p ... --allowedTools ... --output-format json`), concrete paths (`.research/<name>/child_outputs/<id>.md`), and specific thresholds (≥3 subgoals, 8 concurrency, 5–15 min timeouts).	5 / 5
Workflow Clarity	A clearly numbered 0–9 end-to-end sequence with explicit validation checkpoints (double-QA before delivery, a self-check checklist, a mandatory user-confirm gate, and timeout/retry feedback loops), satisfying the batch-operation validation requirement.	5 / 5
Progressive Disclosure	No bundle files exist and the ~270-line body is a monolith with all reference material (basic `claude -p` usage, scheduling/parallel templates) inlined; section headers give internal structure but there are no one-level-deep external references to split the load.	3 / 5
	Total	16 / 20 Passed

Description

100%Weight 40%Scale 1-5

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

A precise, well-structured description that concretely states the capability, gives explicit use cases, and lists natural bilingual trigger terms. Third-person/imperative voice is maintained throughout with no over-claims.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions — 'split a research goal into parallel sub-goals', 'run each via headless `claude -p` subprocesses', 'aggregate results into a polished report file' — covering the full orchestration lifecycle comprehensively.	5 / 5
Completeness	Explicitly answers both 'what' (split/run/aggregate) and 'when' via 'Use for...' and a dedicated 'Triggers:' clause with concrete phrases, meeting the top anchor.	5 / 5
Trigger Term Quality	Comprehensive natural triggers in two languages ('深度调研', 'deep research', 'wide research', '多 Agent 调研', '系统调研') plus use-case phrasing like 'systematic web/document research' and 'competitive or industry analysis'.	5 / 5
Distinctiveness Conflict Risk	Clear niche (multi-agent `claude -p` subprocess orchestration) with distinct bilingual trigger phrases and minimal overlap with other skills.	5 / 5
	Total	20 / 20 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
allowed_tools_field	'allowed-tools' contains unusual tool name(s)	Warning

	Total	15 / 16 Passed

Repository: feiskyer/claude-code-settings
Path: skills/deep-research/SKILL.md
Commit: 6e19e61

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.