Add OpenCode Agent Skill Installation Option to NPX Installer

[grantfox-oss]

Add support to install the Trustless Work skill for OpenCode AI (parity with Claude Code and Gemini CLI) + validate behavior with real integration/verification scenarios

Context

We already have an npm package that users run via:

• Install command: npx @trustless-work/skill
• NPM package page: https://www.npmjs.com/package/@trustless-work/skill

Today, the installer creates/configures the Trustless Work skill for Claude Code, enabling the agent to:

• Guide implementation of Trustless Work products: API, SDK, Blocks
• Verify whether any of those products are implemented correctly in a user's project

We want to extend the installer to also support installing the same skill for OpenCode AI.

---

Goal

Update the npx installer so users can choose the target AI agent and install the Trustless Work skill for:

• ✅ Claude Code (existing)
• ✅ Gemini CLI (existing)
• ➕ OpenCode AI (new)

Critical: The OpenCode skill must have feature parity with Claude and must operate with deep, accurate, and up-to-date knowledge of Trustless Work products and integration requirements.

---

Key Requirement: Skill knowledge must be accurate, detailed, and up-to-date

The installed OpenCode skill must not be generic. It must provide production-grade guidance and verification aligned with current Trustless Work requirements, including (but not limited to):

• Product-specific integration steps for API / SDK / Blocks
• Current required parameters, endpoints, configuration, and conventions
• The latest escrow schema (fields, structure, lifecycle, roles, milestones, disputes, etc.)
• Correct typing expectations (DTOs/interfaces/types), validations, and constraints
• Required conventions (naming, env vars, headers, auth patterns, required payload shape, etc.)
• Verification logic that checks correctness based on the current Trustless Work spec

Source of truth for knowledge: https://docs.trustlesswork.com/trustless-work/

---

Requirements

1) Installer prompt update

• Running npx @trustless-work/skill must show a welcome screen and a target agent selection menu including:
  • Claude Code
  • Gemini CLI
  • OpenCode AI
• Selecting OpenCode AI must trigger the OpenCode-specific installation flow.

2) OpenCode AI skill installation

Implement a new installer path that:

• Creates the skill files/config required by OpenCode AI (format + location per OpenCode’s expected setup).
• Installs the Trustless Work skill with:
  • Implementation workflows: API / SDK / Blocks
  • Verification workflows (detects whether Trustless Work is integrated correctly)

3) Knowledge freshness / update strategy (must-have)

The installer must ensure the OpenCode skill content is updated to match Trustless Work’s latest integration requirements.

Implement at least one of the following strategies:

• Option A (recommended): Add a npx @trustless-work/skill update command that refreshes skill content based on the latest Trustless Work docs/spec.
• Option B: During install, fetch current docs (or a versioned spec snapshot) and embed/reference it so the skill has up-to-date instructions and schemas.
• Option C: Ship a versioned “spec bundle” with a clear upgrade/update path. Warn if the skill/spec is older than the installed package version.

Acceptance expectation: A user who installs the OpenCode skill should get a skill that knows the current parameters, schemas, and typing requirements—not outdated ones.

4) Shared core content (avoid duplication)

Avoid duplicating logic/content across Claude, Gemini, OpenCode:

• Extract shared “Trustless Work knowledge + workflows” into a core module.
• Implement per-agent adapters that map that core into:
  • Claude skill format (existing)
  • OpenCode skill format (new)

5) Graceful behavior

• If OpenCode tooling is not installed or configured, the installer must:
  • Provide a clear message explaining what is missing
  • Provide steps to fix it
• No silent partial installs.

6) Documentation

Update README to:

• Mention OpenCode support
• Show how to install for Claude
• Show how to install for OpenCode
• Explain OpenCode prerequisites
• Include how to update skill knowledge (if an update command is added)

---

Acceptance Criteria

• Running npx @trustless-work/skill shows a menu including Claude Code, Gemini CLI and OpenCode AI.
• Selecting OpenCode AI installs the skill successfully and prints:
  • Where the skill was installed
  • How to invoke/use it in OpenCode
  • Example prompts for:
    • Implement API/SDK/Blocks
    • Verify integration
• The OpenCode skill is not generic and reflects Trustless Work’s real integration needs, including:
  • Correct parameters
  • Correct escrow schema expectations
  • Correct typing/DTO/interface guidance
• Existing Claude installation remains unchanged and works as before.
• README updated.
• Automated tests and/or smoke tests validate:
  • Claude flow still passes
  • OpenCode flow creates the expected files/structure
  • The installed content includes the Trustless Work spec bundle / version reference / update mechanism

---

Mandatory Real-World Validation (must be performed and documented)

Beyond unit tests, the PR must include evidence that the OpenCode skill actually works in realistic scenarios. At minimum:

A) Implementation scenario test (happy path)

Using OpenCode with the installed skill, request an actual implementation, for example:

• “Implement Trustless Work API integration in a NestJS project”
• “Integrate Trustless Work SDK and provide example usage”
• “Add Trustless Work Blocks into a Next.js app with a sample screen”

Expected:

• The agent produces steps/code consistent with current Trustless Work requirements (parameters, schemas, types).
• No outdated instructions.

B) Verification scenario test (error detection)

Intentionally introduce mistakes in a sample project and verify the skill detects them, for example:

• Missing required env vars / config
• Wrong endpoint/parameter names
• Incorrect escrow schema fields
• Type mismatches in DTOs/interfaces
• Wrong payload shape for create/execute flows

Expected:

• The skill identifies the issue(s) clearly and suggests the correct fix aligned with the current spec.

C) Record the results

Add to PR description (or a docs/validation.md) including:

• The prompts used
• The outputs (summaries)
• What errors were introduced
• What the skill detected and how it proposed fixes
• Skill/spec version used during the validation

---

Tests

Run:

• npm test

---

Suggested Implementation Notes

• Add a clean abstraction:
  • installSkill(target: "claude" | "gemini" | "opencode", options)
• Shared modules:
  • core/knowledge (docs/spec bundle, schemas, types, integration checklists)
  • core/workflows (implement API/SDK/Blocks, verify integration)
  • core/verifier (project checks)
• Per-agent adapters:
  • adapters/claude/*
  • adapters/opencode/*
• Add a skillVersion + specVersion field and print it after install.

---

Out of Scope

• Adding support for additional agents beyond OpenCode (other than already planned ones).
• Major CLI refactors unless required to add OpenCode cleanly.

[Chucks1093]

I will extract the shared Trustless Work knowledge into a core module and build an OpenCode adapter on top of it so the skill stays in sync across all three agents without duplicating content. I will have this done the moment it is assigned.

[soma-enyi]

Gm Gm Maintainer!!!!
I would like to work on this issue!!!!

[Macnelson9]

Hi team, I’d like to work on this issue.

I’m a Frontend Developer who thrives at the intersection of UI polish + product usability, and I’ve shipped responsive components/pages using Next.js/React + Tailwind (including interactions, animations, and pixel-accurate Figma builds).

What I’ll deliver

OpenCode Agent Skill Installation implemented to match the design/spec

Responsive behavior (mobile + desktop), no layout breakage

Hover/focus states + accessible semantics

Clean, reusable components and consistent styling

Approach

Review the design/spec and existing code patterns

Build the UI + interactions

QA on multiple breakpoints + open a PR with screenshots/video

If assigned, I’ll start immediately and share progress updates in the PR.

[Joaco2603]

Hi! I’d like to work on this issue.
I have experience working with Node.js CLI tooling and AI agent integrations, so I’m comfortable extending the current npx @trustless-work/skill installer to support OpenCode AI while maintaining parity with the existing Claude and Gemini flows.

My approach would be to introduce a clean installSkill(target) abstraction, extract the shared Trustless Work knowledge/workflows into a core module, and implement an OpenCode adapter that maps the shared logic into the OpenCode skill format. I’ll also implement a spec/version strategy to keep the skill aligned with the latest Trustless Work docs, and include the required real-world validation scenarios.

Happy to follow any repository conventions. If this sounds good, I’d be glad to take it. 🚀

[auracule007]

Hi Maintainers, My name is Promise, a full-stack developer experienced in Node.js, TypeScript, and building CLI tools with npx installers. I understand this OpenCode skill installation issue and can extend the installer to support OpenCode AI with feature parity to Claude and Gemini, implement shared core knowledge modules with up-to-date Trustless Work specs, ensure graceful error handling, and provide documentation with validation scenarios. I'd be glad to take this on, please assign it to me. My ETA is within the 96-hour timeframe.

[estyemma]

Hello Maintainers,
I’m a full-stack and smart contract developer with experience building responsive frontend interfaces, scalable backend systems, and secure blockchain applications.
I’m eager to learn, improve, and collaborate with the community while contributing meaningful value. I would appreciate the opportunity to work on this task.

[edehvictor]

Hello maintainer

Please can I work on this
ETA:3HRS

[Bran18]

Hey team, I’d love to work on this one. I’ve been dealing with similar issues for the past few months and believe I could bring some additional insights to help here