assess-case-test

Assess one free-form request and return a direct proof envelope or focused Plan proposal.

Updated Mar 24, 2026
One-click install
npx skills add https://github.com/skyrocoster/dnd-kids-resources --skill assess-case-test-skyrocoster
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: assess-case-test
Source: https://github.com/skyrocoster/dnd-kids-resources/tree/main/.opencode/skills/assess-case-test
Command: npx skills add https://github.com/skyrocoster/dnd-kids-resources --skill assess-case-test-skyrocoster

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? This Skill standardizes the ASSESS phase of a case-worker workflow, routing a single free-form request to either a direct-change candidate or a complete Plan proposal without editing any files, preventing premature implementation and unbounded investigation. ## Core Features & Use Cases - Bounded Triage: Classifies requests as informational, bug, feature, or refactor within strict budgets of roughly 12 file reads, 4 searches, and 25k input tokens. - Direct Gate: Returns a DIRECT-CANDIDATE envelope with target, evidence, proposed paths, and a proof contract only when every direct criterion is proven. - Focused Plan Gate: Drafts a complete PLAN-CANDIDATE proposal with stages, acceptance criteria, automated gates, ownership globs, exclusions, and stage-grouped compiler evidence. - Use Case: A coordinator hands off a bug report about inconsistent spell card rendering; the Skill compares equivalent rendering sites, checks history, and returns either a direct fix envelope or a staged Plan proposal without touching source files. ## Quick Start Ask the agent to assess this request in the ASSESS phase and return either a DIRECT-CANDIDATE or PLAN-CANDIDATE result without editing any files.

Frequently Asked Questions about assess-case-test

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I assess a request without editing files?▼

Run the ASSESS phase, which triages the request, reads bounded context, and returns either a DIRECT-CANDIDATE or PLAN-CANDIDATE result. The Skill explicitly forbids edits and instructs waiting for resumed authorization after returning a candidate.

When does a request qualify as a direct candidate versus a Plan?▼

A request is DIRECT-CANDIDATE only when it is one atomic logical change with settled behavior, exact paths, known proof, and no design, architecture, API, or migration decisions. Anything needing durable decisions, stages, or coordination becomes a PLAN-CANDIDATE.

What are the investigation limits during assessment?▼

The Skill targets at most 12 file reads, 4 searches, and roughly 25k input tokens. It stops once route, outcome, scope, proof, and compiler-saving evidence are known, and allows at most two repository-wide precedent searches.

Can the assessment phase run repository scripts?▼

Workflow-named repository scripts are invoke-only: use their documented command or --help and read stdout and exit status. The Skill prohibits opening checker, generator, or stage-script source unless changing its behavior.

What happens when evidence contradicts the implementation?▼

For bugs, the Skill compares every equivalent rendering or computation site and uses a bounded history check when contracts, comments, or sibling evidence contradict the implementation, citing that history in the returned evidence.