# Bias-language check

A reusable skill the "Draft a calibrated JD and sourcing brief" recipe pulls in. Add it once and every agent on the team can use it.

Screen a drafted job description for language that narrows or discourages qualified candidates, and fix it — don't just flag it.

## Method
1. Read the full JD draft before it goes anywhere else.
2. Scan for gendered terms, such as "rockstar," "he/she," or aggressive competitive language.
3. Scan for unnecessary requirements: a degree, a years-of-experience minimum, or a tool not actually needed day one.
4. Scan for jargon walls — internal acronyms or insider phrasing an outside candidate can't parse.
5. Rewrite every flagged instance in place with neutral, outcomes-based language.
6. Re-read the rewritten draft once to confirm no flagged pattern remains.
7. Return the clean draft, plus a short list of what was changed and why.

## Rules
- Rewrite every flag before the JD ships — never hand back a flag-only list for someone else to fix.
- Treat the three categories (gendered terms, unnecessary requirements, jargon walls) as the minimum scan, not the ceiling.
- Keep the role's genuine must-haves; cut only what isn't actually required for day one.
- Never soften a real requirement to pass the check — flag it as a policy question instead of silently dropping it.

## Output
The rewritten JD text, plus a change log: each flagged phrase, its category, and its replacement.