private-company-research

AI Berkshire skill: 未上市公司研究：多Agent并行深度研究框架. Source: skills/private-company-research.md.

Quality

49%

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 ./codex-skills/private-company-research/SKILL.md

Quality

Content

70%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 body is a thorough, highly actionable multi-agent research workflow with excellent sequencing and built-in validation/cross-checking feedback loops. Its main weaknesses are length-driven verbosity in the conceptual preamble and a monolithic single-file structure that inlines material which would be easier to navigate if split into reference files.

Suggestions

Trim the "框架特点" and "AI研究偏见自觉" preamble to a few bullet points; Claude already understands information-asymmetry and survivorship bias, so keep only the framework-specific rules (confidence labeling, 第一性原理 fallback mode).

Move the six per-task prompt/description blocks (任务1–任务6) and the final-report template (第七步) into separate reference files under references/ (e.g. references/task-business.md, references/report-template.md) and link to them from SKILL.md, turning the main file into a concise overview.

Either include the referenced tools/financial_rigor.py and AGENTS.md in a bundle directory or replace those references with self-contained inline instructions so the skill is executable without external files that may not exist.

Dimension	Reasoning	Score
Conciseness	The body is mostly efficient action scaffolding (tables, prompt templates, step lists), but the 1082-line length includes explanatory preamble — "框架特点", "AI研究偏见自觉", and the "Codex adapter note" — that teaches concepts Claude already knows and could be trimmed without losing actionable content.	3 / 5
Actionability	It provides copy-paste-ready agent prompt templates, exact tool configs (subagent_type, run_in_background: true), concrete commands (date, python3 tools/financial_rigor.py), and a precise output path, but references tools/financial_rigor.py and AGENTS.md that are not present in any bundle directory, leaving a minor gap.	4 / 5
Workflow Clarity	A clear nine-step sequence (第一步–第九步) with explicit validation checkpoints and feedback loops — 第六步 cross-validation/conflict arbitration, the 反偏见检查, "至少用2个不同来源交叉验证", and "无法获取的信息明确标注数据缺失" — directly matches the anchor for clear sequencing with error-recovery loops.	5 / 5
Progressive Disclosure	Headers organize the document well, but at 1082 lines everything — the six detailed task specifications and the full final-report template — is inlined into a single SKILL.md with no references/ or other bundle files, so content that clearly belongs in separate files is inline and the simple-skill exception (under 50 lines) does not apply.	3 / 5
	Total	15 / 20 Passed

Description

28%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 is essentially a labeled title rather than an actionable capability statement: it conveys the domain but no concrete actions, no natural trigger terms, and no use-when guidance. It is distinguishable from generic skills only because the niche itself is narrow, not because the wording draws a clear boundary.

Suggestions

Rewrite the description to lead with concrete verbs, e.g. "Researches private (pre-IPO) companies by spawning six parallel agents to analyze business model, financials, valuation, competition, risk/governance, and alternative-data signals, then cross-validates and synthesizes an investment report."

Add an explicit trigger clause: "Use when the user asks to research, value, or diligence an unlisted/private company (e.g. Ant Group, Xiaohongshu, SpaceX, Stripe) or mentions pre-IPO valuation, private company due diligence, or 未上市公司研究."

Drop the internal "AI Berkshire skill" label and "Source: ..." attribution from the user-facing description; keep those as metadata or comments instead.

Dimension	Reasoning	Score
Specificity	The description names the domain ("未上市公司研究：多Agent并行深度研究框架") but lists no concrete actions — it is a title plus the meta-label "AI Berkshire skill" and a source attribution, with no verbs describing what the skill actually does.	2 / 5
Completeness	It gives only a vague "what" (a research framework) and entirely omits a "when" / "Use when..." clause, matching the anchor for a vague what with no when rather than the score-3 anchor which requires a clear what.	2 / 5
Trigger Term Quality	Only "未上市公司研究" approximates a phrase a user might say; "AI Berkshire skill" is internal jargon and there are no synonyms, file extensions, or natural trigger terms users would actually utter.	2 / 5
Distinctiveness Conflict Risk	The private-company research niche is somewhat specific and unlikely to collide with most skills, but the description does not stake out distinct triggers and the "AI Berkshire skill" label is ambiguous, so overlap with related research/analysis skills remains possible.	3 / 5
	Total	9 / 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 (1087 lines); consider splitting into references/ and linking	Warning

	Total	15 / 16 Passed

Repository: xbtlin/ai-berkshire
Path: codex-skills/private-company-research/SKILL.md
Commit: be46832

Reviewed: 2 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.