Turn Any Job Description into Interview Intelligence

Where this fits

You're a developer with a first-round interview at a company you barely know. You paste the JD into your agent and ask "拆解 JD" — but instead of a generic summary, you get a structured intel file: every requirement mapped to your own project evidence, company and product research, even a prediction of what the next round will ask. That's the workflow job-intel targets: turning a raw job description into a battle-ready intelligence dossier.

Why agents benefit

• Evidence-driven matching: Every JD line gets matched to concrete evidence from your profile or story bank — if no evidence exists, it says "not found" rather than hallucinating.
• Discrepancy detection: It cross-checks recruiter claims vs. team-level expectations, flagging mismatches that could signal hidden priorities.
• Interviewer profiling: When the next interviewer is known, it builds a profile from public signals (LinkedIn, GitHub, product work) to anticipate their focus.
• Question forecasting: It extrapolates likely next-round questions based on JD gaps and your evidence, so you prepare for the actual conversation, not generic trivia.
• Structured output: Produces a consistent intel.md per job, making it easy to review, share, or feed into downstream skills like interview-script.

Practical scenarios

Preparing for a first-round screen

You share a JD link and your resume. The skill creates jobs/<slug>/intel.md with a full JD-to-evidence table, highlighting where you're strong and where you might get probed. It also researches the company's product and recent news, so your talking points feel current.

After a round, when the next interviewer is known

You just finished a technical screen and the recruiter tells you the next interviewer is a senior engineer who wrote the payment service. job-intel profiles that engineer's public repos and past talks, then forecasts questions around distributed transactions — which you then practice.

When recruiter promises don't match the team's reality

The recruiter said "we use Kubernetes daily," but the team's job postings emphasize bare-metal performance. The skill flags this discrepancy, letting you decide whether to clarify or pivot your narrative.

Add it to your agent workflow

Install with:

npx skills add YunyueLi/greenroom --skill job-intel

Then in your agent, after sharing a JD, trigger it with a prompt like:

Analyze this JD and build my intel file. Create jobs/acme-ai-pm/ and populate intel.md.

The skill will locate your workspace (profile.md, jobs/), create the job file if missing, and generate the six-part intel report.

Compare before adopting

Before committing, evaluate against alternatives: check the skill's GitHub repo for recent commits (maintenance freshness), read the README for quality signals like explicit "do not use" conditions, and test it on your own workspace to see if the evidence-matching resonates with your data. Also consider how it fits with your existing skills — job-intel deliberately avoids answer scripting and debriefing, so it should slot in cleanly.

Why it is worth tracking

With 11 stars and an AGPL-3.0 license, job-intel is early but purpose-built: the author, Yunyue Li, has designed it around a clear principle (evidence over invention) and a strict scope. The skill's value grows as the greenroom workspace pattern matures. Track it if you're building interview-prep agents — the structured intel format could become a reusable standard. See more context at OpenAgentSkill and revisit when the star count climbs or the author adds companion skills.