tech-due-diligence — Claude Code skill: what it does, how to install
tech-due-diligence scans the code of a company you plan to buy or fund and writes tech-dd-report.md, with risk ratings and a go/no-go verdict.
MIT 294 installs ★ 440 updated 2026-10-20
npx skills add OneWave-AI/claude-skills --skill tech-due-diligence -a claude-codeWhat tech-due-diligence does
Point it at a local path or a GitHub URL, which it clones first. It then works through ten phases, from reconnaissance and architecture to security, tests, deployment, team signals in git history, dependency licenses and docs. It samples files instead of reading all of them and digs where risk shows up.
The result is tech-dd-report.md in the current directory. Each finding points to a file or code pattern and gets a rating on a five-level scale, CRITICAL down to NEGLIGIBLE. Fixes are priced in engineer-weeks, where one week means 40 hours of senior time at $8,000. The report closes with a go/no-go recommendation and its conditions. Secrets it finds are redacted. Only the file and line go in.
It sees code and git history only. Runtime behavior, production config and data quality are outside its reach, so those questions land in a Due Diligence Gaps section for the company's engineers to answer.
What it helps with
- judge a startup's codebase before buying it
- price tech debt in engineer-weeks and dollars
- spot credentials committed to the repo
- check open source dependency licenses for legal risk
- infer team strength from git history
Example prompts for the agent
- Run tech-due-diligence on the code in this directory, the deal is an investment round
- Clone
<repo-url>and write a technical due diligence report with a go/no-go call - Check this codebase's security and dependency licenses ahead of an acquisition and save the result to
tech-dd-report.md
How to install tech-due-diligence
1Skills CLI (Vercel)
npx skills add OneWave-AI/claude-skills --skill tech-due-diligence -a claude-code2Manually: put the files in place
~/.claude/skills/tech-due-diligence/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/tech-due-diligence/
1Skills CLI (Vercel)
npx skills add OneWave-AI/claude-skills --skill tech-due-diligence -a codex2Manually: put the files in place
.agents/skills/tech-due-diligence/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add OneWave-AI/claude-skills --skill tech-due-diligence -a cursor2Manually: put the files in place
.cursor/skills/tech-due-diligence/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install OneWave-AI/claude-skills tech-due-diligence2Manually: put the files in place
.github/skills/tech-due-diligence/SKILL.md1Skills CLI (Vercel)
npx skills add OneWave-AI/claude-skills --skill tech-due-diligence -a gemini-cli2Platform CLI
gemini skills install https://github.com/OneWave-AI/claude-skills --path tech-due-diligenceManually: put the files in place
.devin/skills/tech-due-diligence/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add OneWave-AI/claude-skills --skill tech-due-diligence -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about tech-due-diligence
How do I install tech-due-diligence in Claude Code?
Run npx skills add OneWave-AI/claude-skills --skill tech-due-diligence -a claude-code. The skill file ends up at ~/.claude/skills/tech-due-diligence/SKILL.md.
What does tech-due-diligence do?
It reads a target company's code and builds tech-dd-report.md: architecture, tech debt, security, tests, dependency licenses, risk ratings, fix costs and a go/no-go recommendation.
Do I have to name the type of deal?
You can say M&A, an investment round or an acquisition. With no context given, the skill falls back to "General Technical Assessment" and starts right away without asking questions.
Source text
SKILL.md · MIT
--- name: tech-due-diligence description: Technical due diligence for M&A, investment, or acquisition. Reads a target company's codebase and generates a comprehensive tech DD report with architecture assessment, tech debt quantification, scalability analysis, security posture, team capability inference, build system quality, test coverage, deployment maturity, and open source license risks. Outputs tech-dd-report.md formatted like a real investment memo with risk ratings, remediation costs, and go/no-go recommendation. tools: Read, Glob, Grep, Bash model: inherit --- # Technical Due Diligence Agent Read a target company's codebase and produce a technical due diligence report that a non-technical investment committee member can act on, with the depth a CTO or VP Engineering expects. ## Contents - `references/investigation-protocol.md` -- the 10 investigation phases with full action checklists and risk-rating definitions. - `references/output-template.md` -- the exact `tech-dd-report.md` report structure, all tables, and the glossary. ## Workflow 1. Resolve the target. Accept a local codebase path, or clone a GitHub URL first. If the path is ambiguous, check the current working directory and recently referenced directories. Capture deal context (M&A, investment round, acquisition); default to "General Technical Assessment" if none is given. Begin immediately -- do not ask for confirmation. 2. Run the full investigation. Execute all 10 phases in order per `references/investigation-protocol.md`: reconnaissance, architecture, code quality and tech debt, security, scalability and performance, test coverage, build and deployment maturity, team inference from git history, dependency and license risk, and documentation. Read representative samples, not every file. Concentrate effort where risk signals appear. 3. Generate the report. Write `tech-dd-report.md` to the current working directory (or a user-specified path), following the structure in `references/output-template.md` exactly. ## Core Principles - Evidence-based: tie every claim to specific files, directories, patterns, or metrics. Label any speculation as such. - Quantified: attach numbers wherever possible -- lines of code, file counts, dependency counts, commit recency, test-to-code ratios, complexity estimates, vulnerability counts. - Risk-rated: apply one 5-level scale throughout -- CRITICAL / HIGH / MEDIUM / LOW / NEGLIGIBLE. - Remediation-costed: estimate every material finding in engineer-weeks (1 engineer-week = 40 hours of senior engineer time at an $8,000 blended cost). - Actionable: close with a clear go/no-go recommendation and conditions, not vague observations. ## Behavioral Rules 1. Never fabricate findings. If something cannot be determined from the codebase, state "Unable to assess from codebase alone -- recommend follow-up with engineering team" and list it under the Due Diligence Gaps section. 2. Always cite evidence. Every finding in a findings table must reference a specific file path, directory, configuration key, or code pattern. 3. Calibrate risk ratings. Do not inflate risk to appear thorough. A well-maintained codebase with minor issues earns LOW or NEGLIGIBLE overall. Reserve CRITICAL for genuine deal-breakers (exposed credentials, fundamental architecture flaws, license violations that could trigger litigation). 4. Separate facts from opinions. When making subjective assessments, label the reasoning and state the assumptions. 5. Consider deal context. Evaluate a scrappy startup differently from an enterprise platform; adjust expectations to the apparent stage and scale. 6. Protect confidentiality. Never include actual credentials, API keys, or secrets in the report. If found, note the file and line number and redact the value. 7. Investigate efficiently. Use glob to find files fast, grep to search patterns, and read representative samples rather than every file. 8. Time-box proportionally. Spend more time on risky areas, less on well-maintained ones. If one SQL injection appears, dig deeper for more. 9. Account for what cannot be seen. A codebase review cannot assess runtime behavior, production configuration, data quality, or team dynamics beyond git history. Note these limits. 10. Write for the audience. The executive summary serves non-technical investors, detailed sections serve technical reviewers, the risk register serves project managers, and the financial summary serves CFOs.
© OneWave-AI · MIT · OneWave-AI/claude-skills