Skip to content
Forgeplan

Dev Toolkit Plugin

What It Does

Section titled “What It Does”

The dev-toolkit plugin provides language-agnostic development tools for Claude Code and compatible AI agents. It covers the full development loop: research a topic, plan a sprint, build the implementation, audit the result, and restore context when you return.

Installation

Section titled “Installation”
Terminal window
npx skills add ForgePlan/marketplace --plugin dev-toolkit

Commands

Section titled “Commands”
CommandDescription
/auditMulti-expert code audit (4 parallel agents)
/sprint "goal"Adaptive sprint planning with wave-based execution
/recallRestore session context from Hindsight memory
/research "topic"Quick research with Explore agents
/deep-research "topic"Deep multi-agent research (4-7 agents), writes reports
/build "research-dir"Implement from existing research reports
/synthesize "dir1" "dir2"Combine multiple research reports into a unified plan
/do "task"Universal task executor — figures out the right approach
/wave "description"Quick wave-based execution from current context
/write-doc "type" "topic"Create structured documents (RFC, guide, report, ADR)
/team-upLaunch Agent Teams for parallel multi-domain work

/audit — Multi-Expert Audit

Section titled “/audit — Multi-Expert Audit”

Launches 4 agents in parallel, each focused on a different quality dimension:

  1. Logic — correctness, edge cases, race conditions, off-by-one errors
  2. Architecture — SOLID violations, coupling, DRY, naming conventions
  3. Security — OWASP Top 10, injection vectors, auth/authz gaps
  4. Tests — coverage gaps, test quality, missing edge case tests

Each agent reports findings with severity: CRITICAL / HIGH / MEDIUM / LOW.

/audit

Best used before creating a pull request. Pair with /forge-audit from the forgeplan-workflow plugin for methodology-aware auditing.

/sprint — Adaptive Sprint

Section titled “/sprint — Adaptive Sprint”

Scales from a simple task to a full wave-based sprint:

  • 1 task — just executes it directly
  • 3-5 tasks — parallel execution
  • 10+ tasks — wave-based sprint with dependency tracking, progress reporting
/sprint "migrate database to PostgreSQL 16"

The sprint planner breaks the goal into ordered waves, assigns priorities, and tracks completion across waves.

/recall — Context Restore

Section titled “/recall — Context Restore”

Restores session context from Hindsight memory — what you worked on, what was decided, what is pending.

/recall

Use at the start of a new session to pick up where you left off.

/research — Quick Research

Section titled “/research — Quick Research”

Studies a topic using Explore agents. Good for understanding how something works before building.

/research "how does LanceDB handle schema evolution"

/deep-research — Multi-Agent Research

Section titled “/deep-research — Multi-Agent Research”

Deep investigation with 4-7 parallel agents. Writes structured reports to docs/research/. Use before major architectural work.

/deep-research "vector database comparison for embedding storage"

/build — Implement from Research

Section titled “/build — Implement from Research”

Takes existing research reports and creates an implementation plan, then builds it.

/build "docs/research/vector-db-comparison"

/synthesize — Combine Research

Section titled “/synthesize — Combine Research”

Merges multiple research reports into a single unified plan. Useful when you researched several related topics and need one coherent roadmap.

/synthesize "docs/research/auth-options" "docs/research/session-management"

/do — Universal Executor

Section titled “/do — Universal Executor”

Takes any task description and figures out the right approach: research, build, audit, or a combination.

/do "add rate limiting to the API gateway"

/wave — Quick Wave Execution

Section titled “/wave — Quick Wave Execution”

Plans and executes from current chat context without a separate research phase. Good for well-understood tasks.

/wave "refactor the scoring module into separate files"

/write-doc — Document Generator

Section titled “/write-doc — Document Generator”

Creates structured documents using templates and project context. Supports RFC, guide, report, and ADR formats.

/write-doc rfc "LanceDB migration strategy"

/team-up — Agent Teams

Section titled “/team-up — Agent Teams”

Launches parallel Agent Teams for tasks that span multiple domains (e.g., backend + frontend + tests). Each agent works on its domain independently, then results are merged.

/team-up

When to Use

Section titled “When to Use”
  • Brownfield projects — /recall to restore context, /audit to understand quality
  • Large sprints — /sprint with wave-based planning
  • Before major decisions — /deep-research then /build
  • Team coordination — /team-up for parallel multi-domain work
  • Quick tasks — /do or /wave for well-understood changes

See Also

Section titled “See Also”