Use when SynthLabs dataset curation should run as repeatable background HTTP jobs instead of manual UI work.
SkillScore91Maintained
Automatic
◆AI Engineer pack
Last verified 8/21/2026
What it changes
Passed 17 of 19 test cases with the skill, and 6 of 19 without it.
Without skill32%
With skill91%
Checks it passed that the model alone missed
Uses POST /api/jobs/remove-items with dryRun set to true
Polls GET /api/jobs/:id until the task reaches completed status
Identifies DELETE /api/tags/:uid as the route for removing a tag
and 11 more in the full report
13 fixed7 passed either way2 still failing0 broke
Pass-rate lift
+59 pp
Tokens
+59%
Agent turns
0%
Response time
-67%
Model gemini-3.6-flashMeasured 8/21/2026v120 of 22 cases passed with skill · 19 comparable
Tokens, turns and time describe resource use, not a dollar price.
See it in action
One of 12 test cases the skill fixed. Same model, same prompt, with and without the skill.
1 of 12
Prompt
Which HTTP route is used to fetch all organizational tags associated with dataset session `session_alpha_7`?
What was checked, without and with the skill
Checked
Without skill
With skill
Identifies GET /api/sessions/:sessionUid/tags as the route for listing session tags
Missed
Passed
case-12 · gemini-3.6-flash · 8/21/2026 · answers are capped and redacted
What's inside
The instructions your model reads when the skill loads.
SKILL.md161 linessingle file
SynthLabs Curation
Overview
Use this skill for SynthLabs dataset curation that runs through existing HTTP job routes instead of manual UI-first work. This is workflow guidance for an external app; do not assume zorai exposes dedicated SynthLabs tools.
When to Use
Use this skill when:
the task is to autoscore, rewrite, remove items, or migrate reasoning content through background jobs,
the operator wants a repeatable polling loop instead of row-by-row browser actions,
the run should leave a verification trail in the active zorai session, task, or goal notes,
or maintenance checks such as logs stats, tag cleanup, or orphan scans help validate the dataset state.
Do not use this skill when:
the task is row-level manual review, verifier approval, data preview, or other visual inspection,
provider credentials must be entered interactively in the SynthLabs UI,
or the backend is not healthy yet and synthlabs-setup still needs to start or repair it.
Agent Rules
Read current job or session state before starting a new curation job.
Start one specific job, record its jobId, then poll until it reaches completed or failed.
Treat autoscore and rewrite as AI-backed jobs that need a SynthLabs-compatible encrypted apiKey value.
Treat remove-items, migrate-reasoning, orphan checks, and most tag cleanup as non-AI maintenance that does not need encrypted model credentials.
Route row-level manual review, verifier approval, and visual spot checks to synthlabs-ui-operator.
Verification-First Loop
Inspect existing work with GET /api/jobs or session state so you do not start duplicate jobs.
Start exactly one curation route with the narrowest safe payload.
Poll GET /api/jobs/:id until the job is completed or failed.
Review progress, result, and error fields before declaring success.
Update the active zorai session, task, or goal notes with the session ID, job ID, filters used, final counts, and whether UI verification is still pending.
If any rows need human judgment or visual approval, hand that follow-up to synthlabs-ui-operator instead of continuing in blind HTTP mode.