Skill-driven dev workflow for Claude Code, OpenCode, and Codex. writing-plans writes .forge/plan/{slug}.md, then hands off to either in-session execution (executing-plans) or subagent-driven execution (subagent-execution). Includes on-demand brainstorm, debugger, and packaged subagents.
Maps the codebase before planning using GitNexus. Picks the right gitnexus skill based on task type, scans local skills folders for guidelines, and returns a condensed report for the planner.
Runs verification on changed code. Backend → typecheck. Frontend → typecheck + chrome-devtools visual check. Fullstack → both. Reports pass/fail. Delegates deep diagnosis to the debugger skill.
Fetches a Linear ticket via MCP and returns a concise summary. Use when a Linear reference (e.g. ENG-123) is mentioned and ticket context is needed.
Executes a plan by dispatching one fresh subagent per STEP. The controller owns the todo and review; subagents own the edits. Keeps the main conversation light.
Routes dev tasks into the forge planning workflow. Checks whether brainstorm, writing-plans, executing-plans, or subagent-execution should run before any code is touched.
Turns a free-form dev task into a concrete plan on disk. Gathers repo and ticket context, writes a structured plan to .forge/plan/{slug}.md, self-reviews it, then hands off execution.
Use when the user explicitly asks to brainstorm, grill, challenge, compare approaches, clarify domain language, or stress-test a dev idea.
Diagnostic playbook for stuck implementations and failing tests. Classifies symptoms, isolates root causes, and returns minimal fix suggestions. Not for planning or scope expansion.
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
⚠️ Alpha. APIs, skill names, and the plan format may change without notice. Use on throwaway branches until it stabilizes.
Currently biased toward TypeScript projects (typecheck defaults, suggested skills, assumed tooling). PRs welcome to make it language-agnostic.
A skill-driven dev workflow for Claude Code, OpenCode, and Codex: describe → plan → pick mode → execute → test. No orchestrator. Skills trigger on intent and hand off to each other.
Built by developers, for developers.
flowchart LR
User([💬 describe task]) ==> Plan[📝 <b>plan</b><br/><i>writing-plans</i>]
Plan ==> File[(📄 .forge/plan/*.md)]
File ==> Mode{{🎛️ mode?}}
Mode ==>|in-session| Exec[⚡ <b>execute</b><br/><i>executing-plans</i>]
Mode ==>|subagents| Sub[🧩 <b>delegate</b><br/><i>subagent-execution</i>]
Exec ==> Done([✅ test & done])
Sub ==> Done
Plan -.-> Ctx[[🔎 ticket + repo context]]
Exec -.stuck.-> Dbg[[🐞 debugger]]
Sub -.stuck.-> Dbg
classDef main fill:#dcfce7,stroke:#16a34a,stroke-width:2px,color:#111
classDef side fill:#f3e8ff,stroke:#9333ea,stroke-dasharray:3 3,color:#111
classDef pick fill:#fde68a,stroke:#d97706,color:#111
classDef file fill:#e0f2fe,stroke:#0284c7,color:#111
class Plan,Exec,Sub main
class Ctx,Dbg side
class Mode pick
class File file
One path, two execution flavors. Context gathering and the debugger show up only when needed.
| Skill | Role |
|---|---|
using-forge | Entry gate — routes dev tasks into the flow. |
brainstorm | On-demand grill — only runs when explicitly requested. |
writing-plans | Gathers context, writes .forge/plan/{slug}.md. |
executing-plans | Runs the plan in-session. |
subagent-execution | Runs the plan via one subagent per STEP. |
debugger | Diagnostic playbook, loaded on failure. |
| Agent | Role |
|---|---|
ticket-fetcher | Pulls a Linear ticket summary. |
codebase-explorer | Maps relevant files via GitNexus. |
test-runner | Typecheck + optional browser check. |
A single markdown file at .forge/plan/{slug}.md.
{slug} = lowercased Linear ticket id (e.g. eng-123) if present, else a short kebab-case slug (≤ 40 chars, no dates).GOAL, APPROACH, SKILLS TO APPLY, FILES TO CHANGE, STEPS (checklist), TESTS TO UPDATE/ADD, RISKS, OUT OF SCOPE.- [ ] — the executor builds its todo from these.writing-plans overwrites the same file. No v2..forge/ doesn't exist, fall back to .claude/plan/ (Claude) or .opencode/plan/ (OpenCode).There is no orchestrator. The skills chain themselves:
using-forge (meta) which checks whether writing-plans or an executor applies.writing-plans runs → writes the plan → prints the handoff block.1 or 2 → that skill fires and executes.test-runner.brainstorm is available as a direct, on-demand skill when you explicitly want to grill an idea before planning. It is not part of the automatic default path.
You can also invoke any skill directly:
Claude Code:
/write-plan add rate limiter to login endpoint
/execute-plan .forge/plan/eng-482.md
OpenCode:
/skill using-forge add rate limiter to login endpoint
These must be set up manually. The plugin installer does not handle them.
npx claudepluginhub anthonyespirat/forge --plugin forgeHarness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Reliable automation, in-depth debugging, and performance analysis in Chrome using Chrome DevTools and Puppeteer
Comprehensive feature development workflow with specialized agents for codebase exploration, architecture design, and quality review
Feature development with code-architect/explorer/reviewer agents, CLAUDE.md audit and session learnings, and Agent Skills creation with eval benchmarking from Anthropic.
Complete developer toolkit for Claude Code