contextcheck

A Claude Code skill that checks your code changes against a specific company's known engineering patterns, compliance posture, stack conventions, and product philosophy. Uses the research output from /research (local files or gbrain pages).

Designed for work trials, contracted work, takehome assignments, or shipping PRs at a company you've researched. Catches stack/compliance/pattern mismatches before they become review comments.

What it covers

6 alignment dimensions:

1. Stack fit — language, framework, pattern match
2. Compliance & data handling — HIPAA / SOC2 / GDPR alignment if applicable
3. Architectural patterns — connectors, supervisors, error shapes, naming
4. Code quality bar — calibrated to team pedigree (Letta/Jane Street/Perplexity/etc.)
5. Product philosophy fit — does your approach match how they think
6. What would impress — 1-3 specific additions that would stand out

Install

git clone https://github.com/matthewkim323/contextcheck ~/.claude/skills/contextcheck

Use

contextcheck humandelta          # load research/humandelta/ + gbrain
contextcheck                     # auto-detect company from repo metadata
contextcheck --file src/x.ts     # check a specific file
contextcheck --save              # append report to research/<company>/ALIGNMENT_HISTORY.md
contextcheck --quiet             # only print warns/criticals, skip passes
contextcheck --all               # chain /prodcheck + /seccheck + /contextcheck

Or just say "does this fit their stack" / "check against company context".

Pairs with other skills

skill	run when
/research	first — to build the context contextcheck needs
/prodcheck	general production readiness
/seccheck	security audit
/contextcheck	company-specific alignment
/prwrite	last — generate the PR body

Or just run contextcheck --all to do the full pre-submit gauntlet.

License

MIT.