HR0101/company-analysis

Company Analysis & Interview Preparation Master (Company Analysis for Claude)

Conducts tailored company analysis based on primary sources (IR materials, pitch decks, integrated reports). User reviews are treated as unverified reference information. When interview preparation is requested, it uses 5-axis patterns to generate competitive comparisons, a ~60-second self-introduction, motivation statements, in-depth responses, hypothesis-driven reverse questions (standard 10), and expected interview questions from the panel (standard 100).

Reference Files and Script Locations

If installed as a plugin, they are located at:

• ${CLAUDE_PLUGIN_ROOT}/skills/company-analysis/references/
• ${CLAUDE_PLUGIN_ROOT}/skills/company-analysis/scripts/

If placed directly at ~/.claude/skills/company-analysis/ (or .claude/skills/company-analysis/ / the Claude Desktop app), use the references/ and scripts/ directories in the same directory as this SKILL.md. All subsequent path references are relative to this skill directory.

---

1. Confirm the Request's Purpose and Necessary Conditions

• Start with Input UI: For requests starting with "I want to start with the input UI," "I want to analyze by selecting items," or skill activation without a specified target company or purpose (e.g., /company-analysis or /company-analysis:company-analysis alone), follow references/input-ui.md and actually call an available question tool (e.g., AskUserQuestion in Claude Code). Have the user input/select the company name, purpose, and necessary axes using the arguments generated by scripts/input_ui.py. Do not consider displaying a JSON or Markdown menu as "showing the UI." In environments without the question UI (e.g., Claude Desktop / Web chat), switch to short text questions.
• Company Research / IR Analysis / Industry/Competitor Comparison: Focus the analysis on the industry, company type, and the specific points requested. Requests with only a company name should proceed as a holistic company research project. Confirmation of desired role, interview stage, or applicant category is not required. Do not generate motivation statements or reverse questions. Use only the necessary axes from the 5 below.
• Application & Interview Preparation: Set the following 5 axes based on information provided by the user. Only if missing information significantly impacts the deliverables, use the question UI in references/input-ui.md to confirm the necessary conditions. Do not ask for already answered items; supplement the industry/company type with reasonable estimates if possible. For an "Automatic" request, proceed with explicitly stated assumptions. Do not fabricate the user's background, achievements, or degree of motivation.

Axis	Pattern ID & Options	Detailed Reference
① Industry Pattern	it-saas-web (IT/SaaS/Web/AI/SIer)consulting (Strategy/General/IT Consulting/Think Tank)manufacturing (Manufacturer: Electrical/Automotive/Semiconductor/Chemical/Food/Pharma, etc.)trading (General Trading Company / Specialized Trading Company)finance (Finance: Banking/Securities/Insurance/Cards/Asset Mgmt/FinTech)media-ad-entertainment (Advertising/Media/Entertainment/Games)real-estate-infra-logistics (Real Estate/Developer/Construction/Infrastructure/Logistics)hr-retail-service-healthcare (HR/Retail/EC/Healthcare/Education/Services)	references/industry-patterns.md
② Desired Role Pattern	sales-cs (Corporate Sales/Individual Sales/IS/FS/Customer Success)planning-marketing-bizdev (Corporate Planning/Business Planning/Product Planning/Marketing/BizDev)pdm-pm-consultant (PdM/PM/Consultant/Presales)software-ai-engineer (Web/Mobile/AI/Data/SRE/Security)rd-manufacturing-eng (R&D/Mechanical-Electrical-Chemical Design/Production Engineering/Quality Assurance)corporate (Finance & Accounting/HR/Legal/IR & Public Relations/Information Systems)	references/role-patterns.md
③ Interview Stage	casual-intern-ob (Internship/OB/OG Visit/Casual Meeting)es-resume (Document Screening/ES/Resume Creation)first-interview (1st Interview: For HR/Young to Mid-level On-site Staff)second-interview (2nd / Mid-stage Interview: For On-site Section Manager/Department Manager)final-interview (Final / Executive Interview: For President/Director/Business Unit Head)offer-comparison (Post-offer/Offer Meeting/Multiple Company Comparison)	references/interview-stage-patterns.md
④ Applicant Category	new-grad (New Graduate Job Hunting) / mid-career (Mid-career Job Change) / second-new-grad (Second New Graduate)	references/interview-stage-patterns.md
⑤ Company Type	listed-enterprise (Major Japanese / Listed) / mega-venture (Mega-Venture) / startup-unlisted (Unlisted Startup) / foreign-affiliated (Foreign-affiliated)	references/secrets-and-methodology.md

Immediate Blueprint Generation (Using Scripts)

Once the interview preparation patterns are determined (or to view a list), you can generate a blueprint using the included script. For company-research-only requests, there is no need to run this script to fill in unnecessary interview axes.

In the following commands, scripts/ is resolved relative to the directory containing this SKILL.md (when installed as a plugin: ${CLAUDE_PLUGIN_ROOT}/skills/company-analysis/). Unknown or ambiguous specifications will cause an error, so re-specify using valid IDs.

# Display a list of all patterns
python3 scripts/pattern_selector.py --list

# Output a strategy blueprint for a specified pattern (automatically resolves Japanese keywords and abbreviations)
python3 scripts/pattern_selector.py \
  --company "Target Company Name" \
  --industry "IT" \
  --role "Engineer" \
  --stage "2nd Interview" \
  --candidate "Mid-career" \
  --archetype "Mega-Venture" \
  --competitors "Competitor A, Competitor B"

---

2. Core Procedure for Company Analysis (Applying the Methodology)

Before starting the analysis, be sure to grasp the principles in references/secrets-and-methodology.md and load only the references for the required industry, role, and interview stage. For company research, references for role and interview stage are not required.

Step 1: Fact Collection from Primary Sources (Web Search & Scrutiny of Provided PDFs)

• Reading Attached PDFs/Local Documents (Utilizing the Read Tool):
  • If the user specifies or attaches PDFs or files such as securities reports, financial results, medium-term management plans, integrated reports, job postings, or resumes/curriculum vitae, read them directly using the Read tool (for PDFs exceeding 10 pages, use the pages parameter to specify a range, reading up to 20 pages at a time).
  • If the latest information is needed, search official IR pages, EDINET, PR TIMES, etc., using the WebSearch / WebFetch tools.
• Listed Companies (listed-enterprise / mega-venture):
  • Securities Report (Business overview, segment-wise revenue/profit, challenges to address, business risks, employees' average age, tenure, average annual salary)
  • Financial Results Presentation (Highlights of the latest full-year/quarterly performance, KPI trends, year-on-year change factors)
  • Medium-term Management Plan (Quantitative targets for 3-5 years, growth investment areas, business portfolio reform)
  • Integrated Report (CEO message, value creation process, human capital data)
• Unlisted Startups (startup-unlisted):
  • Pitch Decks (Company Deck / Culture Deck), Funding announcements on PR TIMES (Round, total funding amount, underwriting VCs/corporate investors), Official Gazette financial statements, Founder interviews / Wantedly.
• Foreign-affiliated Companies (foreign-affiliated):
  • Global Headquarters Annual Report / Form 10-K and the Mission scope of the Japan office (Is it a sales/marketing hub, or does it have independent planning/R&D/development authority?), Job Description (JD) requirements breakdown.

Step 2: Financial, Cash Flow & Human Capital Gap Diagnosis

When financial figures, employee data, or review metrics are obtained, you can organize quantitative indicators, the relationship between CF sign classification and amounts, and entered review metrics using scripts/financial_analyzer.py. Align the fiscal year, currency units, and scope (individual/consolidated) for input, and note the source separately in the report.

python3 scripts/financial_analyzer.py \
  --company "Target Company Name" \
  --revenue 1500 --prev-revenue 1350 --gross-profit 600 --operating-profit 180 --net-profit 120 \
  --total-assets 2000 --equity 1000 --employees 3500 \
  --operating-cf 220 --investing-cf -140 --financing-cf -50 \
  --avg-age 36.2 --avg-tenure 11.4 --avg-salary 850 \
  --ow-overall 3.7 --ow-growth20s 3.5 --ow-compliance 4.1

• Reconciling Official Announcements (Integrated Reports, Career Site) with On-site Sentiment (OpenWork, etc.):
  • Check the posting date, number of reviews, role, and department of user reviews against the scope of the official information. Organize any discrepancies as hypotheses to be verified.
• CF Interpretation: The 8 sign classifications represent cash inflows/outflows. Evaluate the purpose and soundness of investments and financing after confirming the amounts, beginning/ending cash balances, repayment schedules, breakdown, and multi-year trends. Zero is not classified as either positive or negative.

Step 3: Three Transformations to Turn "Research" into "Success"

Only when application/interview preparation is requested, transform the collected facts into the following three deliverables (see Section 5 of references/secrets-and-methodology.md for details):

1. Competitive Comparison Matrix → "Why This Company":
   • Line up 2-3 rival companies in the same industry and create a comparison table of their business portfolios, customer bases, profit structures, medium-term plan investment targets, and differences in culture.
2. Company Challenge Hypothesis × Applicant's Strengths → "Why Me":
   • Based on the "challenges to address" in the securities report or the "focus themes" in the medium-term plan, hypothesize the bottlenecks faced by the on-site team of the desired role (role), and connect how the applicant's experience can be leveraged.
3. Primary Source Fact + On-site Hypothesis → "Hypothesis-Validating Reverse Questions":
   • Design reverse questions perfectly tailored to the interview stage using the syntax [IR/Official Fact] + [My On-site Hypothesis] + [Question Tailored to Interviewer's Position] (1st interview = on-site junior staff/HR, 2nd interview = on-site manager/section head, final = president/executive).

---

3. Selecting and Generating the Output Format

For company research, IR analysis, and competitive comparison, use "Template 0: For Company Research/IR Analysis" from references/output-templates.md, focusing on the requested points. For application/interview preparation, use the interview preparation templates from the same file or scripts/report_generator.py. The scripts generate worksheets with blank fields, so supplement the content with reviewed materials before delivery. If only the blank template itself is requested, it's acceptable to deliver it as-is.

Clearly distinguish between the following two types of questions, and do not confuse the input UI, headings, or quantity settings. Use the standard quantities common to both the full version and the last-minute version.

• Reverse Questions (Applicant → Company/Interviewer): Standard 10 questions. Specify using reverse_question_count and --reverse-question-count N. Construct questions based on primary source facts + on-site hypothesis + a query to confirm with the other party. Do not include unsubstantiated assumptions.
• Expected Questions (Company/Interviewer → Applicant): Standard 100 questions. Specify using interview_question_count and --interview-question-count N. Structure around ~10 themes such as self-introduction, motivation/company understanding, role/expertise, achievements/problem-solving, collaboration, failure/learning, values, career choices, work style, and post-entry contributions. Phrase these as questions the interviewer would ask the applicant, tailored to the desired role, applicant category, interview stage, and the applicant's background. For high-priority questions, add the interviewer's confirmation intent and a key answer outline. Do not fabricate unconfirmed background or achievements as answers.

Prioritize the number specified by the user via the input UI. If zero questions are requested for a type, omit that type entirely. At delivery, confirm the number of questions per type, sequential numbering, and lack of semantic overlap. In the last-minute version, keep the short main text and separate the two question sets, clearly labeling the numbers for the expected questions (to prioritize preparation) and the reverse questions (to ask the company on the day). These question counts do not apply to company research or offer comparison.

For the full and last-minute versions of interview preparation, follow references/self-introduction.md to automatically generate a ~60-second self-introduction using available conversation history and provided background materials. Separate from the motivation statement, create a script for a spoken introduction covering: current affiliation → representative experience/achievements → strengths → closing, aiming for 250-350 characters. If sufficient background is already available, do not ask for re-entry. Only if information is missing, confirm the necessary items in a single question using the input UI. Do not use past chats that you cannot access or information you cannot confirm as the applicant's own achievements. When outputting via the script, save the completed script to a temporary file and pass it via --self-introduction-file.

# Generate the Full Company Analysis & Interview Success Master Sheet (--mode full)
python3 scripts/report_generator.py \
  --company "Target Company Name" --industry "Trading" --role "Planning" --stage "2nd Interview" \
  --candidate "New Grad" --archetype "Major Japanese" --competitors "Competitor A, Competitor B" --mode full

# Generate the 5-Minute Pre-Interview Cheat Sheet (--mode cheatsheet)
python3 scripts/report_generator.py \
  --company "Target Company Name" --industry "IT" --role "Sales" --stage "Final Interview" --mode cheatsheet

# Generate the Post-Offer / Company Comparison Scorecard (--mode offer-comparison)
python3 scripts/report_generator.py \
  --company "First Choice Company" --competitors "Concurrent Offer A, Concurrent Offer B" --mode offer-comparison

---

4. Mandatory Quality Guardrails

1. Prohibition of Fabricating Numbers and Facts (No Hallucination):
   • Quantitative data such as revenue, profit, average salary, and paid leave rate must only be stated as verified facts (with the fiscal year and source specified). Do not fill in missing items with estimates. Distinguish between "unconfirmed" (uninvestigated, source not verified), "unobtainable" (access to materials failed, viewing restrictions), and "undisclosed" (confirmed as not disclosed). Include the verification method and reference materials. Do not claim a fact is undisclosed just because a search failed to find it. Even commonly reported themes in user reviews should be treated as "review trends/hypotheses" until corroborated by primary or other reliable sources.
2. Differentiating Perspective for Interview Preparation:
   • If the interviewer's position/role or the job description is known, prioritize that information over estimation based on the interview number. The following are general guidelines.
   • First Interview: Emphasize "concise, conclusion-first answers (45-60 seconds), basic understanding, honesty and collaborative ability." Avoid overly academic management criticism.
   • Second Interview: Emphasize "clear comparison with competitors, demonstrable work reproducibility in the desired role, and hypothesis-building from an on-site manager's perspective."
   • Final Interview: Emphasize "commitment to the company as a top choice, alignment of corporate vision/mission/values (VMV) with the applicant's formative experiences, and a long-term vision for 5-10 years."
3. Avoiding a One-Sided "Fan Letter":
   • Present not only the company's strengths but also risk items from the securities report, structural challenges in the industry, and points requiring confirmation derived from user reviews--clearly separating verified facts from hypotheses--in a constructive manner.