company-analyzer — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited company-analyzer (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Produces a company analysis organized as Story × Logic × Judgment — a deliberate departure from SWOT-style templated reports. The three layers map to how humans actually understand a business: the narrative arc makes it memorable, the economic logic makes it rigorous, and the forced judgment makes it useful.
The benchmark for style and depth is 《晚点 LatePost》-class industry writing (see references/style-benchmark.md for an annotated example). If the output reads like a translated McKinsey slide with English jargon sprinkled in, it failed — even if all the facts are correct.
This is the hardest part to get right and the easiest to fail. Before writing any section, hold the following rules in memory.
Default: use AskUserQuestion to confirm three things before any web search:
跳过条件: 用户在请求里已经明确指定了公司 + 目的 + 深度(或者深度从上下文清楚),直接进入 Phase 1,不要为了"按流程"打断用户. 但如果任一维度歧义(公司名有两个候选 / 没说目的 / 没说深度),必须问.
按公司生命周期选 ONE. 下列权重是叙事重心,不是字数机械切分.
references/financial-analysis.md.派 3 个子 agent 并行. 每个 agent 负责一层,收到 prompt 后要做的第一件事是独立核实父 agent 传过来的初始事实——创始人履历、年份、关键数字这些,如果和公开资料冲突,以独立核实为准,不要照搬父 agent 的叙述。父 agent 也会出错。
references/story-patterns.md.references/business-logic.md、references/moat-types.md、和 references/industry-structure.md.references/anti-patterns.md first. 在 Story + Logic 落地后运行.主 agent 把三层编织成一篇文章,不是三段拼接. 语言风格参照 references/style-benchmark.md 的《晚点》示例:平白中文、整句论断作标题、段间用"但问题在于"/"对比而言"/"所以"做过渡. 不保留英文术语标签作为 section header(不要出现 "### Moat Analysis" 这种东西).
量化配额(中文字符数),低于下限视为研究深度不够,高于上限视为灌水.
(注:这是 synthesis 后最终产出字数,不是三层草稿相加;单层草稿会在合成时被压缩和穿插.)
详见 references/anti-patterns.md. 短清单:
references/style-benchmark.md — 《晚点 LatePost》风格标注示例(v2 新增)references/industry-structure.md — 如何做行业结构分析(v2 新增)references/story-patterns.md — 五种叙事形态(成熟公司可简化使用)references/business-logic.md — 收入机制 + 商业性质 + 管理层分析references/moat-types.md — 7 类护城河 + 监管/IP/资本(用作思考工具,输出时用中文自然语言)references/financial-analysis.md — 上市公司财务 4 步读法references/anti-patterns.md — 详细负面清单 + 重写对照templates/report-template.md — 建议骨架(标题按一句话论断写)~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.