problem-solving

Orchestrates specialized agents for research, root cause analysis, architecture decisions, and validation with persistent artifacts.

Updated Jul 2, 2026
One-click install
npx skills add https://github.com/geekatron/jerry-claude-plugin --skill problem-solving-geekatron
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: problem-solving
Source: https://github.com/geekatron/jerry-claude-plugin/tree/main/skills/problem-solving
Command: npx skills add https://github.com/geekatron/jerry-claude-plugin --skill problem-solving-geekatron

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Complex problems often get shallow, one-shot answers that disappear when the chat context fills up. This Skill routes your request to one of nine specialized agents (researcher, analyst, architect, critic, validator, synthesizer, reviewer, investigator, reporter) and persists every output as a markdown artifact so findings survive session compaction and build a project knowledge base. ## Core Features & Use Cases - Specialized Agent Routing: Keyword-based selection across nine agents covering research, 5 Whys root cause analysis, ADR creation, constraint validation, cross-document synthesis, code/security review, incident investigation, and status reporting. - Mandatory Artifact Persistence: Every agent writes structured output to docs/ directories (research, analysis, decisions, reviews, investigations, reports) with L0/L1/L2 audience levels. - Adversarial Quality Mode: Creator-critic-revision cycles with a 6-dimension weighted scoring rubric and minimum 3 iterations for C2+ deliverables. - Use Case: Ask "Research PostgreSQL vs MongoDB, then create an ADR for our choice" and the orchestrator chains ps-researcher into ps-architect, producing a cited research document and a Nygard-format ADR. ## Quick Start Describe your problem in natural language, such as asking to analyze why your CI tests are flaky using 5 Whys, and the appropriate agent will activate and persist its findings to the docs directory.

Frequently Asked Questions about problem-solving

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

FAQPage Schema
How do I choose the right problem-solving agent for my task?▼

Describe your need in natural language and the orchestrator matches keywords to agents: "research" or "explore" selects ps-researcher, "why" or "root cause" selects ps-analyst, "ADR" or "decide" selects ps-architect, and "investigate" selects ps-investigator. You can also name an agent explicitly.

How do I create an architecture decision record with AI agents?▼

Request an ADR in natural language, such as "Create an ADR for choosing Redis as our caching layer." The ps-architect agent produces a Nygard-format ADR persisted to docs/decisions/ with L0, L1, and L2 explanation levels.

Can multiple problem-solving agents work together on one task?▼

Yes, agents chain sequentially or in parallel through the orchestrator. A typical workflow runs ps-researcher, then ps-analyst, then ps-architect, then ps-validator, with each agent reading upstream artifacts. Agents cannot spawn other agents directly per the P-003 constraint.

Where are agent outputs saved and do they survive context compaction?▼

All outputs are persisted as markdown files under docs/ subdirectories such as research, analysis, decisions, reviews, and investigations. This mandatory persistence rule (P-002) ensures artifacts survive session compaction and accumulate as project knowledge.

What is the adversarial quality mode and when does it apply?▼

It is a creator-critic-revision cycle where ps-critic scores deliverables using a 6-dimension weighted rubric. C2+ deliverables require a minimum of 3 iterations and a composite score of at least 0.92 before passing the quality gate.

When should I not use the problem-solving skill?▼

Use other skills for simple workflow coordination (/orchestration), NASA systems engineering reviews (/nasa-se), transcript parsing (/transcript), standalone adversarial scoring (/adversary), threat modeling (/eng-team), penetration testing (/red-team), or work item tracking (/worktracker).