Skip to content

SherpaBuild the right thing. Prove it works. Show why it matters.

A portable AI mentor with client-specific skill and command integrations for hackathon teams.

Layered mountain ridges with a winding trail to a small summit flag
curl -fsSL https://github.com/Anshkaliya18/Sherpa/releases/download/v2.0.0/install.sh | bash

Sherpa 2.0.8 adds generated client distributions and a managed installer. Check release availability and compatibility notes before installing.

A good guide helps you choose a route, checks the weather, and leaves the decisions with you. Sherpa brings that rhythm to a hackathon build.

Route to the summit ​

  1. Base CampDiscover the problem and compare ideas/sherpa-suggest
  2. Camp 1Choose a feasible plan/sherpa-plan
  3. Camp 2Build in small, reviewable steps/sherpa-build
  4. Camp 3Verify behavior and review risks/sherpa-verify
  5. High CampShape a clear, evidence-backed demo/sherpa-demo
  6. SummitPrepare a complete submission/sherpa-submission

Find a command ​

28 commands

  • /sherpa-statusCORE

    Show evidence-backed project status.

  • /sherpa-nextCORE

    Identify the single highest-priority next action.

  • /sherpa-helpCORE

    List implemented sherpa commands and limitations.

  • /sherpa-initCORE

    Initialize project-local sherpa state safely.

  • /sherpa-phaseCORE

    Show current lifecycle phase and evidence.

  • /sherpa-configCONFIGURATION

    Show a redacted project configuration overview.

  • /sherpa-config-checkCONFIGURATION

    Validate project configuration and secret protections.

  • /sherpa-suggestINTELLIGENCE

    Discover, evaluate, and red-team idea candidates; human selection required.

  • /sherpa-evaluateINTELLIGENCE

    Evaluate and compare candidate ideas with evidence.

  • /sherpa-compareINTELLIGENCE

    Compare candidate ideas using the existing idea-evaluation workflow.

  • /sherpa-redteamINTELLIGENCE

    Challenge project assumptions and readiness using the red-team workflow.

  • /sherpa-optimizeINTELLIGENCE

    Prioritize score-improving work against evidence and criteria.

  • /sherpa-judge-meINTELLIGENCE

    Simulate judge perception using the judge-simulation workflow.

  • /sherpa-repoENGINEERING

    Inspect repository evidence as part of project planning.

  • /sherpa-architectureENGINEERING

    Plan architecture from the selected idea and repository evidence.

  • /sherpa-techstackENGINEERING

    Assess technology choices within the project plan.

  • /sherpa-planENGINEERING

    Create an implementation-ready project plan.

  • /sherpa-buildENGINEERING

    Execute an approved implementation task using the existing implementation workflow.

  • /sherpa-verifyENGINEERING

    Verify implementation through the existing testing and security workflow.

  • /sherpa-testENGINEERING

    Test implemented scope using the testing and security workflow.

  • /sherpa-securityENGINEERING

    Review security using the testing and security workflow.

  • /sherpa-uxENGINEERING

    Review product usability and accessibility readiness.

  • /sherpa-accessibilityPRODUCT QUALITY

    Review accessibility within the existing UX-review scope.

  • /sherpa-demoFINALIZATION

    Prepare an evidence-backed judge-oriented demo plan.

  • /sherpa-auditFINALIZATION

    Run the existing final-audit workflow after its prerequisites are met.

  • /sherpa-submissionFINALIZATION

    Prepare and verify submission materials without submitting.

  • /sherpa-postmortemFINALIZATION

    Run the final-phase evidence-based retrospective.

  • /sherpa-researchINTELLIGENCE

    Research competitors, markets, technologies, and evidence with verifiable sources.

Why teams use Sherpa ​

Evidence-led

Separate facts, assumptions, and unknowns before turning them into a plan.

People keep the gates

Human approval stays explicit for consequential choices and implementation steps.

State travels with the project

Working notes and generated evidence live in the repository’s .sherpa/ directory.

Compatibility is stated plainly

Integration files and runtime verification are reported as separate things.

Agent compatibility ​

AgentIntegration filesInstaller scopeRuntime
Claude Code`.claude-plugin/`, `.claude/skills/`, `skills/sherpa/`npm installer or pluginNOT VERIFIED
Gemini CLI`gemini-extension.json`, `.agents/skills/`shared skill directory or extensionNOT VERIFIED
Qwen Code`qwen-extension.json`, `.agents/skills/`shared skill directory or extensionNOT VERIFIED
OpenCode`.opencode/commands/`, `.agents/skills/`, package metadatanpm installer or shared skillNOT VERIFIED
Kilo Code`kilo.jsonc`, `.kilo/`, `.agents/skills/`shared skill directoryNOT VERIFIED
GenericManually load root `SKILL.md` if host supports instruction filesHost-dependentHost-dependent

See the full agent compatibility notes for scope and evidence.

Keep exploring ​

GitHub repository · MIT License · version 2.0.8

Build with evidence. Keep the decisions yours.