skill-verifier-candidates-v1

SkillDev tools

Mine latest-generation skill_invocation evidence into verifier candidate rows (one row per skill x archetype) for the Skill Verifier Factory.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the skill-verifier-candidates-v1 skill

What this skill tells your AI

The instructions your AI receives, as published by jayfarei/opentraces in src/opentraces/workflow_templates/skill-verifier-candidates-v1/SKILL.md and read by ahel’s review.

Project the Trace Index skill_invocation read side into verifier candidate rows: one row per (skill, archetype) describing how strong the trace-grounded evidence is for building a verifier of that archetype, which contract markers are most deficient, whether a leakage-safe Dtrain/Dsel/Dtest split is feasible, and the verifier-creator interview (with evidence-inferred defaults).

This is the canonical workflow surface over opentraces.consumers.verifier_factory.mine_verifier_candidates; the opentraces workflow verifier-factory consumer reads these rows (and emits the packages). Recommendations are advisory: no verifier is generated or promoted without explicit human approval.

Scope

  • project (optional): restrict mining to one project slug.
  • skills (optional list): restrict to specific skills (default: every skill that has a registered archetype).
  • min_usable_episodes (optional int, default 30): data-gap threshold.

Output

Newline-delimited JSON rows validated against schemas/row.schema.json (schema_version: opentraces.skill_verifier_candidates_row.v1).

Signals

GitHub stars
99
Forks
6
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
skill-verifier-candidates-v1
Source
github.com/jayfarei/opentraces