deep-research-agent

深度研究技能，用于系统性调研与分析。适用于研究、调研、竞品分析、市场分析或结构化信息收集等需求。

Quality

67%

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

Fix and improve this skill with Tessl

tessl review fix ./app/skills/deep-research-agent/SKILL.md

Quality

Content

56%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.

The skill has a genuinely clear incremental workflow with strong validation gates, and its shell commands match the bundled scripts. Its main weaknesses are pseudocode tool calls for the central save/update operations, redundant repetition of the cleanup rules, and a broken reference to a templates directory that isn't shipped.

Suggestions

Replace the `file_write`/`file_update` pseudocode with real, executable operations (e.g. the Write/Edit tools or a bundled script call), including a concrete search_pattern example instead of "{研究背景说明}".

Consolidate the placeholder-cleanup rules (§4 更新规则, §5.1, §5.3) into the single §5 validation section and drop the duplicated init command line.

Either ship the `deep-research-agent/templates` directory and list its template files, or inline the template selection logic so the reference is not dangling.

Dimension	Reasoning	Score
Conciseness	The stepwise structure is mostly efficient, but the placeholder-cleanup rule is repeated three times (§4 更新规则, §5.1, §5.3 plus the §5.4 checklist) and the init command is stated twice back-to-back ("```bash ... ```" followed by "用法： ..."). Consolidating these into one validation section would tighten it.	3 / 5
Actionability	The two bash commands (init_research.py, md_to_html.py) are concrete and match the real bundled scripts, but the core research loop relies on non-existent `file_write`/`file_update` tool calls with template placeholders like "{研究背景说明}" embedded in search_pattern — pseudocode rather than executable guidance.	3 / 5
Workflow Clarity	The six-step sequence is clearly laid out with a gated validation phase (§5.1–§5.4) and an explicit final checklist before generating research_data.json. It falls short of 5 only because the error-recovery loop is implicit — there is no 'if a check fails, fix and re-verify' instruction.	4 / 5
Progressive Disclosure	The body is well-sectioned and its script references (scripts/init_research.py, scripts/md_to_html.py) are real, one-level-deep, and match the bundle, but the twice-referenced `deep-research-agent/templates` directory does not exist in the bundle, and the body explicitly declines to enumerate it ("无需在此列出具体模板名称"), leaving a dangling, undiscoverable reference.	3 / 5
	Total	13 / 20 Passed

Description

61%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.

The description has a clear, explicit use-when clause with several natural trigger terms, but the capability statement itself is vague — it never says what the skill actually produces (incremental markdown/HTML reports with cited sources). Tightening the 'what' and adding distinctive deliverables would lift both specificity and distinctiveness.

Suggestions

Replace the abstract "系统性调研与分析" with concrete capabilities, e.g. '按章节增量生成带引用来源的研究报告（report.md / report.html / research_data.json）' (systematically build cited research reports section by section).

Add distinctive deliverables or trigger phrases (e.g. '调研报告', '行业分析', '查资料', '文献综述') to reduce overlap with generic search/analysis skills.

Mention the outline-confirmation-first workflow in the description so users can distinguish it from a plain search skill.

Dimension	Reasoning	Score
Specificity	The description names the domain ("深度研究技能") but its only actions are the abstract "系统性调研与分析"; the remaining phrases (竞品分析、市场分析、结构化信息收集) are application domains rather than concrete capabilities like report generation or source citation.	2 / 5
Completeness	Both parts are present: a 'what' ("用于系统性调研与分析") and an explicit 'when'-equivalent ("适用于研究、调研、竞品分析、市场分析或结构化信息收集等需求") with concrete triggers. It falls short of 5 because the 'what' is generic and omits concrete deliverables such as the incremental report, research_data.json, or HTML output.	4 / 5
Trigger Term Quality	It includes several natural trigger phrases users would say — 研究/调研 (a synonym pair), 竞品分析, 市场分析, 结构化信息收集 — though common variations such as 查资料, 行业分析, or 文献综述 are missing, so coverage is good but not comprehensive.	4 / 5
Distinctiveness Conflict Risk	The research/market-analysis triggers are somewhat specific but could overlap with generic web-search or analysis skills; only 竞品分析 gives it a partial niche, and there are no distinctive artifacts or file types to separate it from similar research skills.	3 / 5
	Total	13 / 20 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: ZHangZHengEric/Sage
Path: app/skills/deep-research-agent/SKILL.md
Commit: 81bddde

Reviewed: 7 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.