Company Recon

A Claude Code skill that investigates a company before you interview there, using what employees say instead of what the careers page says.

Paste a job description or a company name. It comes back with a full dossier: how long the money lasts, who the founder actually is and what happened at their last company, who left the leadership team in the past 18 months and whether anyone replaced them, which companies the staff came from and where the alumni went, and what the reviews say once you weight the people who already left.

Then it gives you two verdicts, not one.

Install

In Claude Code, as a plugin. This is the one to use: it brings the skill and the /company-recon-init command.

/plugin marketplace add jainulabudeenm/company-recon
/plugin install company-recon

Or from npm, if you want the skill in Claude.ai or anywhere outside Claude Code. This installs the skill on its own, without the slash command.

npx company-recon

Add --project to install into the current repo instead of your home directory, so you can commit it and share it with a team.

Use

> Here is a job description, is this worth my time?
> [paste]

On first run it interviews you once about your target band, your stability bar, and your hard filters, and saves the answers to ~/.claude/company-recon/profile.md. Every finding afterward is scored against your constraints rather than a generic idea of a good job.

Re-run /company-recon-init any time your situation changes.

The two verdicts

Most job advice tells you whether to take a job. That throws away most of the value of a hiring process.

Should you join? Strong yes, proceed with caution, or would not join. It has to cite the runway rating and the mentorship rating by name, so a shaky company cannot hide behind a good pitch.

Should you interview anyway? This is the useful half. Four tiers:

• Target. You want it. Run the full campaign.
• Leverage. You would not join at their band, but a written offer here anchors the offer you actually want. Optimize for a fast, high number rather than fit.
• Reps. You will not join and the offer moves nothing, but the loop trains one named weakness: presenting your work, the whiteboard round, negotiating out loud. Name the weakness or it is just calendar noise.
• Skip. Rare. Nothing to learn, or an exposure risk you will not take.

A company you would never work for is still worth mining. The skill will not let you delete one without first asking what you could get out of it.

Two rules it will not break

Never invent a number. Runway, headcount, pay bands: it cites the source and the date, or it writes "not disclosed". A confident wrong number is worse than an admitted gap.

Never delete a company. Every dossier ends in both verdicts, so a no is never the end of the conversation.

Timing and exposure

It schedules leverage interviews backwards from when your target company's offer is expected, so both numbers land in the same two week window. An offer that arrives a month early anchors nothing.

It also flags when applying somewhere risks reaching your current employer through shared investors, ex-colleagues on the panel, or a recruiter working both sides. It does not decide that for you. It makes sure you know before you apply.

Sources

Financial filings, Crunchbase, layoff trackers, LinkedIn People and Insights, Glassdoor, Blind, Reddit, and news coverage. It weights people who have left a company over people still there, who have an incentive to be diplomatic. Filing sources are listed per region, covering India, the United States, and the United Kingdom.

The profile

~/.claude/company-recon/profile.md, in your home directory rather than a project folder, because it holds compensation figures and exit intent. That should be impossible to commit to a repo, not just unlikely. Plain markdown, edit it whenever. See profile.example.md for the useful level of detail.

License

MIT