Works with the AI Agents you already use
Post-Training Data Pack: Framing, Curation, Preference Labels
Gets fine-tuning data training-ready: problem framing, governance, curation, preference labeling, and synthetic data, with pass/fail gates.
$5
Post-Training Data Pack: Framing, Curation, Preference Labels
Example session with this skill installed
We scraped 20k support tickets to fine-tune a support bot. Some have customer emails in them and we also want to generate synthetic tickets for rare issues. What do we need to fix before training?
- Read your context and instructions
- Compiled the post-training data
Situation: no baseline yet; PII present; rare intents under-covered. Blocking: (1) M2 - redact emails and other PII, confirm you're allowed to train on ticket content, and dedupe before splitting; (2) M3 - split by customer or time so the same thread can't land in train and test. Plan: M1 baseline on a 400-ticket held-out set per main intent; M3 intent taxonomy with per-slice minimums; M13 synthetic tickets only for intents under the minimum, filtered and human-spot-checked, capped as a share of the mix. Next action: run a PII scan and count tickets per intent.
Connects securely to your tools. The creator never sees your data.
About this skill
Bad data is the most common reason fine-tunes fail, and it's usually found after the GPU bill. This skill makes your agent act as the data lead before anything gets trained.
What it does
- Frames the project first: baseline, success metric, stop criteria, and whether training is even the right call.
- Checks data is usable: provenance, licenses, consent, PII and secret redaction, train/test leakage, benchmark contamination.
- Curates: taxonomy, deduplication, slice balance, splits, versioning, dataset cards.
- Designs preference labeling: rubrics, pairwise vs binary labels, annotator calibration, agreement and bias audits.
- Plans synthetic data and distillation without model collapse.
- Audit mode: share a dataset sample, card, or labeling guide and get pass / fail / missing per gate with fixes.
Part of the Post-Training series. The full Post-Training Gatekeeper bundles all four packs.
Plans and audits data work; it does not run scanners or process files itself. Not legal advice on licensing or privacy law.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 17 days ago
Permissions
No special permissions declared or detected