Audit recruiting artifacts (JD, scorecard, interview questions, rubrics) for bias risks: coded language, requirements that disproportionately filter out qualified candidates, rubrics with adjective-only anchors, panel composition issues. Outputs specific edits.
How this skill is triggered — by the user, by Claude, or both
Slash command
/ciromaciel-recruiting:bias-mitigation-checkThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Pure reasoning skill. Audits recruiting artifacts for bias risks and proposes concrete edits. Not a moralistic review — a quality review focused on outcomes (better hires, wider qualified pools, fewer false negatives).
Pure reasoning skill. Audits recruiting artifacts for bias risks and proposes concrete edits. Not a moralistic review — a quality review focused on outcomes (better hires, wider qualified pools, fewer false negatives).
Built for: Recruiters who want their process to actually be fair, not performative about it.
/intake produces scorecard + JD (always — pre-publication audit)/report if diversity mode shows drop-offRead whichever artifacts user points to:
job-description.mdscorecard.mdinterviews/<NN>-<type>.mdassessment-template.mdCoded language check
Flag and rewrite:
Requirement bloat check
For each "Required":
Research backing: women apply at 100% match; men at 60%. Every unnecessary "required" cuts your qualified pool disproportionately. Trim to 4-6 hard musts.
Salary transparency check
Range present? If not + jurisdiction requires it (CA, NY, CO, WA, EU pay transparency) → REFUSE to publish.
Imagery / inclusion signaling check
If JD includes language like "we welcome people of all backgrounds" — only retain if the rest of the JD matches (else feels performative). If process has zero accommodations language → add at minimum: "Need accommodation in the interview process? Reply to this email."
Competency check
Level rubric check
For each competency, are the 4 levels described with observable behavior or adjectives?
Bad: "4 - Excellent communication" Good: "4 - Adapts framing in real-time to match listener's context"
If adjectival, force rewrite.
Knockout check
For each knockout question, can you defend it as truly disqualifying?
Question quality check
Bias trap check per question
Panel composition
If panel is documented:
Generate clients/<cliente>/jobs/<job-slug>/bias-audit.md:
# Bias mitigation audit — [Vacancy] | [Date]
## Summary
- Artifacts audited: [list]
- Issues found: [N total — H high, M medium, L low]
- Status: [READY / NEEDS-REVISION / BLOCKER]
## Findings
### HIGH (blocker — fix before publishing/using)
1. **[Issue title]** in `[file:line]`
- Current: "[quote of problematic text]"
- Why this matters: [1 sentence — outcome-focused, not moral]
- Suggested rewrite: "[new text]"
### MEDIUM (fix soon)
[same format]
### LOW (consider for future)
[same format]
## Quick wins (do these in 10 min)
- [Specific action 1]
- [Specific action 2]
- ...
## Deeper changes (require discussion with hiring manager)
- [Action with rationale]
Bias mitigation done right = wider qualified pool + fewer false negatives + better hires. If a "fix" makes the process worse (longer, less signal, less comparable), it's not a fix. Push back.
npx claudepluginhub ciro-maciel/ciromacielos --plugin ciromaciel-recruitingProvides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.