About help
help is an Agent skill router that matches a user's situation to a suitable skill and hands off using a gather-match-handoff process. Its routing patterns cover engineering, research, browser tasks, and creative work.
Key features
- 01Matches skills to situations
- 02Uses gather-match-handoff routing
- 03Covers engineering, research, and creative tasks
Use cases
- 01Routing a browser interaction request
- 02Routing an Agent system design task
- 03Routing a content editing request
Install
Copy the command and run it in a terminal.
$
npx skills add https:/After installing, type/helpin Claude Code to confirm it loaded.
Bundled files
References and scripts installed along with SKILL.md.
references/gstack-sources.mdreferences/agent-browser/LICENSEreferences/agent-browser/references/commands.mdsources.mdreferences/agent-systems/references/evaluate-agents/references/comparison-method.mdreferences/deslop/evals/acceptance.jsondevelopment.jsonreferences/deslop/references/pattern-catalog.mdreferences/deslop/references/humanizer/LICENSEreferences/deslop/references/no-ai-slop/eval.mdLICENSEreferences/deslop/references/unslop/LICENSEreferences/distill-skill-to-model/references/pipeline-plan.mdreferences/focus-content/evals/acceptance.jsondevelopment.jsonreferences/llm-wiki/references/original-idea.mdreferences/manage-project/references/ai-agent-processes.mdreferences/marketing/LICENSEreferences/marketing/references/context.mdreferences/marketing/references/pricing/references/research-methods.mdreferences/marketing/references/seo-audit/references/technical-checks.mdreferences/product-management/references/ai-opportunity-assessment.mdcatalog.mdconversation.mddemand-discovery.mdplatform.mdquestion-model.mdspecifications.mdreferences/skill-routers/references/canonical-design.mdwriting-patterns.mdDetails
From the project READMEHelp
Match the user's situation to the skill best suited to handle it.
Read the shared routing guide and use its gather → match → handoff procedure with these patterns.
| Situation pattern | Skill |
|---|---|
| Substantial work has returned incomplete and needs finishing against every original requirement, or the user explicitly requests unlazy completion discipline. Use ordinary domain workflows for routine work, verification-only requests, or wording such as "be thorough" without a completion-recovery need. | unlazy |
| The requested outcome is operating a website, filling a form, extracting rendered data, or capturing a page through browser interaction. Software fixes, QA reports, and design research retain their engineering owner, which may use this browser workflow for evidence. | agent-browser |
| A TypeSafe System One or Jev integration needs designing, implementing, reviewing, or debugging against current model behavior and documented jagged edges. | typesafe-ai |
| An agent skill router needs explaining, designing, implementing, or auditing. | skill-routers |
| LoRA training is already chosen: a source skill needs distilling into a small adapter with a sealed comparison against that skill. | distill-skill-to-model |
| Grok Bot capabilities, skills, routines, roster, autonomy rules, handoffs, or templates need creating or changing. | grok-bot |
| A persistent Git-backed LLM wiki needs setting up, ingesting sources, answering queries, linting, or compacting. | llm-wiki |
| An agent system's harness, context, memory, tools, interaction, coding loop, evaluation, training strategy, learning loop, or multi-agent architecture needs designing or improving; or an existing skill needs compiling into a cheaper runtime representation without an already-chosen LoRA target. | agent-systems |
| A multi-task project needs a durable record to coordinate work, blockers, asynchronous results, review, or resumption. | manage-project |
| A plan, estimate, proposal, roadmap, or strategy needs testable forecasts, probabilities, and response rules. | predictive-planning |
| Product capabilities, features, journeys, requirements, releases, positioning, customer evidence, or AI opportunities need defining or maintaining in a durable product record. | product-management |
| Customer research, pricing or packaging decisions, persuasive marketing-page copy, organic-search audits, or lifecycle email sequences need a focused workflow. | marketing |
| A workflow needs a source-grounded callstack trace, state-machine projection, or both, without executing it. | workflows |
| Existing content needs narrowing to the three primary questions it should answer, with every unrelated section, component, or item deleted. | focus-content |
| Technical documentation needs creating, restructuring, or auditing around a reader's learning, task, lookup, or understanding need. | technical-documentation |
| Existing prose needs editing or auditing for AI-writing patterns while preserving its meaning and voice. | deslop |
| Established software laws need explaining, applying to a decision, or assessing against evidence from a concrete case. | software-laws |
| A Git repository's commit history needs transforming into an evidence-grounded animated story, picture book, or illustrated narrative. | software-craft |
| A concept, prompt, design direction, name, narrative, or other creative artifact needs developing for novelty and taste. | elicit-llm-creativity |
| Engineering work on executable software, infrastructure, libraries, services, or single-workflow agent skills needs investigation, scoping, repository mapping, design, planning, a claimable work queue, implementation, review, verification, delivery, coordination, or knowledge transfer. | software-craft |
Comments
Markdown supported