sn-deep-research

用于用户请求深度研究、系统性研究、竞品分析、方案对比、趋势分析或事实核查时。**遇到以下任一情况就主动使用本 skill，不要自行搜几条就回答**：①用户出现触发词：深度研究 / 深度调研 / 深入研究 / 全面研究 / 系统研究 / 调研 / 调查 / 尽调 / 行业研究 / 市场研究 / 竞品分析 / 政策研究 / 技术研究 / 趋势研究 / 事实核查 / 写一份研究报告 / 调研报告 / 深度报告 / research / deep research；②请求需要跨多来源取证、多维度对比、交叉验证才能给出可靠结论；③用户要求产出报告、白皮书、行业分析或尽调文档；④话题涉及最新政策/市场/产品/价格/法规，需要系统核查。明确要求核验来源的单点事实可走 quick；无核验要求的简单常识问答不使用。模糊或宽泛的"研究/了解一下 X"也优先触发。仅不用于：一句话摘要、已给定单一来源的整理、纯文字润色改写。

Quality

83%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

88%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 highly actionable, well-gated orchestration skill with explicit validation feedback loops and copy-paste-ready commands. Its main weakness is token redundancy from repeating the payload-first-line rule across every stage rather than relying on the §2.2 contract.

Suggestions

Rely on the §2.2 contract for the mandatory '先读取 {plugin_role_dir}/<role>.md 并严格遵守' first line instead of repeating it verbatim in all 12 §5 stage payloads, trimming meaningful tokens.

Consider splitting the §5 stage library into a per-stage reference file (or the existing agents/*.md) and keeping SKILL.md as the dispatch index, which would shorten the inlined monolith and lift progressive_disclosure.

Consolidate the repeated path-resolution placeholder notes (§2.1) and the per-stage path lines into a single defined-variables block referenced by shorthand.

Dimension	Reasoning	Score
Conciseness	Dense operational prose with no padding about concepts Claude already knows, but the ~630-line body carries redundancy — e.g., the §2.2 'every payload starts with 先读取 {plugin_role_dir}/<role>.md' rule is then repeated verbatim in all 12 §5 stage payloads — so it sits below the lean anchor-5 bar.	4 / 5
Actionability	Fully executable, copy-paste-ready guidance throughout: concrete bash (mkdir/openssl rand/python3 validate_plan.py with exact flags), exact file and schema paths, and complete payload templates with named fields — matching the anchor-5 example.	5 / 5
Workflow Clarity	Explicit sequenced pipelines per mode (§4.2 quick/normal/heavy), validation gates with ok:true/false routing (§5.4, §5.9), a failure/retry table with criteria/route/limits (§4.3), and feedback loops (validate→fix→re-validate, supplement-research cycles) — a clear anchor-5 match.	5 / 5
Progressive Disclosure	Good structure with a top reading map and six numbered sections, plus one-level-deep references to real bundle scripts (validate_plan/evidence/outline.py, source_snapshot.py) and external schemas/role files; held below 5 because the full §5 stage library is a large inlined block in a long single file rather than split out.	4 / 5
	Total	18 / 20 Passed

Description

77%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 trigger-rich, boundary-explicit description that excels at "when to use" but is weaker on stating the skill's concrete capabilities. The aggressive lean-to-trigger stance slightly raises overlap risk with generic search skills.

Suggestions

Add a brief 'what' clause stating the skill's concrete actions in third person (e.g., 'Orchestrates expert sub-agents to gather multi-source evidence, cross-validate findings, and render cited research reports') so capability is not only implied by triggers.

Soften or scope the '模糊或宽泛的"研究/了解一下 X"也优先触发' directive, since broad single-phrase triggers overlap with general search/research skills and raise conflict risk.

Lead with a single crisp capability sentence before the numbered trigger conditions, so the 'what' precedes the 'when' and reads as more than a trigger list.

Dimension	Reasoning	Score
Specificity	Names the research domain and many concrete task/deliverable types ("竞品分析、方案对比、趋势分析、事实核查", "报告、白皮书、行业分析或尽调文档"), but states no concrete action verbs for what the skill itself does (orchestrate, validate, render) — those live only in the body.	3 / 5
Completeness	The "when" is explicit with four numbered conditions and concrete trigger phrases, but the "what" is indirect — it lists research types rather than stating the skill's mechanism, so it sits below the anchor-5 bar that requires both crisply stated.	4 / 5
Trigger Term Quality	Comprehensive natural trigger coverage with extensive synonyms in Chinese (深度研究/调研/调查/尽调/竞品分析/事实核查/写研究报告…) plus English (research/deep research), matching the anchor-5 example of broad synonym coverage.	5 / 5
Distinctiveness Conflict Risk	A clear niche (deep, multi-source, report-producing research) with explicit exclusions ("仅不用于：一句话摘要、已给定单一来源的整理、纯文字润色改写"), but the directive "模糊或宽泛的'研究/了解一下 X'也优先触发" creates minor overlap risk with general search skills.	4 / 5
	Total	16 / 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
skill_md_line_count	SKILL.md is long (638 lines); consider splitting into references/ and linking	Warning

	Total	15 / 16 Passed

Repository: OpenSenseNova/SenseNova-Skills
Path: skills/sn-deep-research/SKILL.md
Commit: 61b60bb

Reviewed: 8 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.