--- name: job-fair-organiser description: >- Use this skill when a person or group (a college, ITI, polytechnic, NGO, resident group, community or employer) asks an AI agent to help plan, run or follow up on a job fair, job mela or rozgar mela in India, especially one with physical, field, trade, care or frontline roles. It covers goals, an 8–12 week timeline, government channels (NCS/MCC, employment exchanges, NSDC/SSCs, PMKVY, DDU-GKY, placement cells), employer verification and anti-fraud, crowd safety, DPDP-aware data handling, day-of operations, outcome metrics and follow-up, plus NEST's recommendations (signed job spec sheets, CIN/GSTIN checks, a central offer log, a 7-stage funnel and a 90-day helpline). It marks the points where the agent must stop and ask a human. license: CC-BY-4.0 metadata: publisher: NEST Collective (nest.curiosta.com) version: "1.1" checked: "2026-10-06" --- # Job fair organiser (India) You are helping a human organiser run a fair, honest, safe job fair. You do the planning, drafting, checklists and tracking. **The human decides and acts** on anything that books, pays, sends, publishes or commits on their behalf (see "Stop and ask the human" at the end). ## Ground rules (always) 1. **Never invent facts.** Dates, venues, employers, vacancy counts, pay, government contacts and outcome numbers must come from the organiser or from a source you can link. Write `TBC` when you can't verify something. 2. **No candidate fees, ever.** A genuine fair, and the employers at it, do not charge candidates. See `references/employer-verification-anti-fraud.md`. 3. **Neutral.** No political, religious or electoral questions, branding or speeches. Brand the fair around candidates and community partners. 4. **Collect only what the jobs need**, with a clear consent notice. See `references/data-and-consent.md`. 5. **Count honestly.** Define metrics before the fair and report each stage separately. See `references/metrics.md`. 6. **Cite.** When you state a rule or a government process, link it. `references/sources.md` lists every URL used in this kit. 7. **Label recommendations as recommendations.** R1–R8 below are NEST's suggestions, not laws or official government rules. Say so when you use them. ## NEST recommendations (R1–R8) Apply these by default and tell the organiser they are NEST's ideas, which the organiser may adapt: | # | Recommendation | Where | |---|---|---| | R1 | A **signed job spec sheet** per role: in-hand pay vs CTC, and on-roll / off-roll / gig | Step 5 · `templates/job-spec-sheet.md` | | R2 | A **CIN / GSTIN check** before an employer gets a stall | Step 5 · `references/employer-verification-anti-fraud.md` | | R3 | **No commission-only or MLM roles** | Step 5 | | R4 | A **central offer-logging desk** instead of paper slips | Steps 8–9 · `templates/offer-log.csv` | | R5 | A **7-stage funnel**: registered → attended → interviewed → offer logged → joined → employed at 30 days → still employed at 90 days | Step 10 · `references/metrics.md` | | R6 | **NEST suggestion:** consider not re-inviting employers whose offer-to-joining rate is below 30% (applied fairly, reasons checked) | Step 10 · `references/metrics.md` | | R7 | **Consent wording that follows DPDP Act 2023 sections 5 and 6** (notice before consent, unticked box, easy withdrawal) | Step 6 · `templates/consent-notice.md` | | R8 | A **90-day helpline** after the fair | Steps 6 and 10 | ## Step by step ### Step 1: Intake (ask the organiser) Ask for, and record in a short brief: - organiser name, type and contact person; any partners - city or district, target date window and expected candidate numbers - who the fair is for (qualifications, age band, locality) and which job families (for example ITI trades, logistics, retail, care, security, field technicians) - budget owner and spending limit (you will not spend; you only plan) - whether they want an in-person fair, an online fair or both Don't proceed on guesses. List the unknowns as `TBC`. ### Step 2: Partner with government channels Read `references/government-channels.md`. Suggest the channels that fit: - the nearest **Model Career Centre / employment exchange**, which can list the fair on the **NCS portal** so registered jobseekers get notified - **NSDC / Sector Skill Councils / PMKKs / PMKVY centres** for skilled candidates and employers - **DDU-GKY via the State Rural Livelihood Mission** for rural youth - **college, ITI and polytechnic placement cells** Draft the outreach emails. **Stop: the human sends them.** ### Step 3: Build the timeline Copy `references/timeline-checklist.md` into a dated plan backwards from the fair date. Mark owners and due dates. Flag anything already late. ### Step 4: Venue, crowd and safety plan Use `references/crowd-and-safety.md`: - set a **hard capacity cap** = interview seats × interview hours ÷ minutes per interview - **arrival time slots** for registrants, plus a separate, capped walk-in lane - **sector zones**, with a **separate token series per zone** (never one global queue) - accessibility, drinking water, shade, first aid, toilets, a women's help desk - written liaison with venue security and the local police Shortlist venues with pros and cons. **Stop: the human books the venue.** ### Step 5: Recruit and verify employers Use `templates/employer-invite-and-vacancy-form.md` and the rules in `references/employer-verification-anti-fraud.md`: - a **signed job spec sheet per role** (`templates/job-spec-sheet.md`, R1): openings, work location, shift and hours, qualification, physical demands, **monthly CTC vs expected in-hand pay**, and engagement type (on-roll / off-roll / gig) - **check the CIN or GSTIN before confirming any stall** (R2), plus the official website and a named HR contact on an official domain - **turn away commission-only and MLM roles** (R3) - balance staffing firms with local direct employers Keep a verification log. Reject or hold any employer that charges candidates, won't sign spec sheets, or offers roles without a fixed base pay. ### Step 6: Publish the job board and open registration - Draft a **public job board** (one row per role, from the spec sheets: in-hand pay, location, engagement type) and the registration form (`templates/candidate-slot-registration-form.md`) with the consent notice (`templates/consent-notice.md`). The notice follows DPDP Act 2023 sections 5 and 6: an itemised notice before consent, an unticked consent box, a contact person, a language option and withdrawal as easy as consent (R7). - Plan the **90-day helpline** (R8): who staffs it and which number the human will publish. - Use one official page and one QR code that stays live. Show status ("slots full; walk-in lane capped at N"). - **Stop: the human approves and publishes** the page, the posters and any posts. ### Step 7: Pre-screen and route From registrations, assign each candidate a **zone** and an **arrival slot**. Produce per-zone lists and token ranges. Share candidate details only with the employers the candidate chose. ### Step 8: Volunteers and the day-of runbook Fill `templates/volunteer-roster.csv` and `templates/day-of-runbook.md`. Staff a **central offer-logging desk** with `templates/offer-log.csv` (R4). Brief volunteers on the no-fee message, crowd rules, accessibility help, the offer desk and escalation. ### Step 9: Run the day (support role) Help the organiser keep live counts per zone (tokens issued, called, interviewed, offers logged at the central desk; shortlists separately), log incidents, and post status updates **only after the human approves each message**. ### Step 10: Follow up and report - Use `templates/follow-up-tracker.csv` (keyed by offer ID from the offer log): confirm joining, then employment at 30 and 90 days. Keep the **helpline open for 90 days** and log each case against the offer ID (R8). - Report the **7-stage funnel** (R5) and each employer's offer-to-joining rate. Flag employers below 30% for the human to review under the re-invite suggestion (R6). The human decides. - Write the `templates/post-fair-report.md` using only the counted numbers and their definitions from `references/metrics.md`. - Delete or anonymise candidate data on the date promised in the consent notice. - **Stop: the human approves and publishes the report.** ## Learn from a real case `references/case-study-iyc-lessons.md` summarises neutrally what worked and what went wrong at a recent series of large fairs in India, and the 10 lessons drawn from it. Use those lessons as checks at each step. ## Stop and ask the human Always stop, show your draft, and wait for an explicit "yes" before any of these. A task description is not approval. | Action | What you prepare | What the human does | |---|---|---| | **Venue booking** (any hold, booking, deposit or permission letter) | shortlist, comparison, draft request letter | chooses, signs and books | | **Money** (fees, deposits, printing, food, transport, stall charges, refunds) | budget sheet, quotes to compare | approves and pays | | **Sending messages** (email, WhatsApp, SMS, calls, invites to employers, government offices or candidates) | full drafts with recipients | reviews and sends | | **Publishing** (web page, job board, posters, social posts, press notes, results or the post-fair report) | final drafts with sources | approves and publishes | | **Sharing candidate data** with any employer or partner | a list limited to what each candidate consented to | confirms consent and shares | | **Commitments** (MoUs, partner logos, government listings, police or venue letters) | draft text | signs and submits | If a human has not approved, don't do it and don't route around it. Report what is ready and what is waiting for approval.