cat-eval

name	cat-eval
description	Routes work to the correct SuperGrok role inside the eval domain (rubrics, judges, harnesses). Activates when selecting a eval specialist, browsing eval agents, or invoking /cat-eval. Differentiator: category index only — does not perform eval implementation work.
argument-hint	<role or goal within eval>
user-invocable	true
metadata	{"short-description":"Navigate eval SuperGroks (6 roles)","category":"eval","tier":"core","sg_id":"sg-nav-eval","binary_id":"opgrok.sg.cat-eval","version":"2.0.0","leslie_gate":"v1","kind":"navigator","intent":"Route work to the correct eval SuperGrok role.","purpose":"Index 6 eval specialists for rubrics, judges, harnesses.","intent_tags":["eval","navigator","route","index"],"path":"core/skills/eval/SKILL.md","call":"/cat-eval"}

Eval SuperGrok Navigator (/cat-eval)

Agent Identity: Benedict-086cc36e0cdb08c767912c8b424c44f16cb43eaff785d485c45ae27bef49cb68

Core Mandate / Invariants

• Domain: rubrics, judges, harnesses.
• This skill selects a specialist; it does not implement the domain work.
• Prefer a single primary SuperGrok; secondary agents only for mesh handoff via /opgrok.

Procedural Workflow

1. Parse the goal into a role method (build, review, map, harden, document, integrate, …).
2. Map method → role among: audit, forge, scout, seal, smith, trace.
3. Invoke /eval-<role> or open core/skills/eval/<role>/SKILL.md.
4. If multi-agent composition is required, escalate to /opgrok.
5. Emit: SELECTED: /eval-<role> with a one-line justification.

Constraints & Gotchas

• Do not perform heavy implementation under this navigator.
• Do not invent roles outside the catalog.
• Ambiguous goals: choose the highest-risk path role (anvil/crux) or scout first.

Definition of Done

• One primary call path named (/eval-<role>).
• Intent/purpose of that skill fits the goal.
• Optional secondaries listed only if mesh is required.

Optional Tool Surface

• Registry: core/skills/_framework/REGISTRY.json
• MCP catalog: core/skills/_framework/MCP_CATALOG.json

References

• core/skills/_framework/NAVIGATION.md
• core/skills/_framework/ENHANCEMENT_PROTOCOL.md
• Roles: /eval-audit, /eval-forge, /eval-scout, /eval-seal, /eval-smith, /eval-trace