SF Compound Engineering Plugin v3.1.0-beta.3

Instruction-Based Compound Engineering for Salesforce — a multi-platform plugin (Claude Code, Cursor, Codex, and 9 other AI coding tools) where each iteration becomes smarter than the last through institutional knowledge capture and parallel persona dispatch.

> V3 is a skills-first architecture. Commands were retired — every entry point is now a skill that auto-routes from natural-language phrases via its description frontmatter, with direct slash invocation (/sf-<name>) still supported. See CHANGELOG.md for the migration guide from v2.x.

***

The Compound Engineering Loop

  ┌─ HUMAN (taste) ─┐        ┌──────── AI (in the loop) ────────┐        ┌─ HUMAN (taste) ─┐
   Ideate · Brainstorm  →  Plan(40%) · Deepen · Work(20%) · Review · Resolve  →  Polish  →  Compound  →  Repeat

  • Ideate     — decide what's worth building (/sf-ideate)
  • Brainstorm — explore requirements through collaborative dialogue (/sf-brainstorm)
  • Plan       — research & design using 60 skills + parallel research personas (/sf-plan)
  • Deepen     — enhance the plan with section-level parallel research (/sf-deepen)
  • Work       — implement with pre-research + system-wide test checks (/sf-work)
  • Review     — parallel persona dispatch across 60 specialist personas (/sf-review)
  • Polish     — taste pass via /sf-polish: SLDS2/UX, accessibility (WCAG), copy (UI surfaces only)
  • Compound   — capture learnings to docs/solutions/, personas, skills, CLAUDE.md (/sf-compound)

> The sandwich. Humans own the two ends — Ideate (what's worth building) and Polish (does it feel right) — the "bread". The AI runs the middle "filling" in the loop. As models get better at execution, human attention concentrates where machines are still weak: taste and judgment.

> All nine core entry points (/sf-ideate, /sf-brainstorm, /sf-plan, /sf-deepen, /sf-work, /sf-review, /sf-polish, /sf-compound, /sf-lfg) are skills in V3 — they auto-route from natural-language phrases via their description frontmatter, and direct slash invocation continues to work.

Above the loop, /sf-strategy maintains an optional repo-root STRATEGY.md (target problem, approach, users, key metrics, tracks) that sf-ideate, sf-brainstorm, and sf-plan read as grounding when it exists.

Each iteration starts smarter because learnings compound into docs/solutions/, personas, skills, and CLAUDE.md.

> Principles. This plugin is opinionated. Seven principles — preserve the quality ceiling, verifiability, stay in the loop, the spec is the artifact, taste over typing, agent-native docs, outsource thinking not understanding — govern every skill and every review. See PRINCIPLES.md. Each core workflow skill declares which principles it enforces.

***

Quick Start

Claude Code (Native)

# Add as a Claude Code plugin marketplace
/plugin marketplace add https://github.com/gellasangameshgupta/salesforce-compound-engineering-plugin

# Install
/plugin install sf-compound-engineering

Other AI Coding Tools (11 platforms)

# GitHub Copilot
bunx @gellasangameshgupta/sf-compound-plugin install sf-compound-engineering --to copilot

# Cursor
bunx @gellasangameshgupta/sf-compound-plugin install sf-compound-engineering --to cursor

# Windsurf
bunx @gellasangameshgupta/sf-compound-plugin install sf-compound-engineering --to windsurf

# Gemini CLI
bunx @gellasangameshgupta/sf-compound-plugin install sf-compound-engineering --to gemini

# OpenCode
bunx @gellasangameshgupta/sf-compound-plugin install sf-compound-engineering --to opencode

# Codex
bunx @gellasangameshgupta/sf-compound-plugin install sf-compound-engineering --to codex

# Kiro
bunx @gellasangameshgupta/sf-compound-plugin install sf-compound-engineering --to kiro

# Factory Droid
bunx @gellasangameshgupta/sf-compound-plugin install sf-compound-engineering --to droid

# Pi
bunx @gellasangameshgupta/sf-compound-plugin install sf-compound-engineering --to pi

# OpenClaw
bunx @gellasangameshgupta/sf-compound-plugin install sf-compound-engineering --to openclaw

# Qwen Code
bunx @gellasangameshgupta/sf-compound-plugin install sf-compound-engineering --to qwen

# Auto-detect and install to all detected tools
bunx @gellasangameshgupta/sf-compound-plugin install sf-compound-engineering --to all

# Sync from current directory to all detected tools
bunx @gellasangameshgupta/sf-compound-plugin sync

What Gets Converted

| Platform | Commands | Agents | Skills | MCP Config | | ------------ | -------------------------- | ------------------------------ | ---------------------------- | -------------------------- | | Copilot | .github/skills/ | .github/agents/.agent.md | .github/skills/ | copilot-mcp-config.json | | Cursor | Sync only | Sync only | .cursor/skills/ (symlinks) | .cursor/mcp.json | | Windsurf | workflows/.md | skills//SKILL.md | skills/ | mcp_config.json | | Gemini | .gemini/commands/.toml | .gemini/skills/ | .gemini/skills/ | settings.json | | OpenCode | commands/.md | agents/.md | skills/ | opencode.json | | Codex | prompts/.md + skills/ | skills//SKILL.md | skills/ | config.toml | | Kiro | .kiro/skills/ | .kiro/agents/.json | .kiro/skills/ | mcp.json | | Droid | commands/.md | agents/.md | skills/ | mcp.json | | Pi | prompts/.md | skills//SKILL.md | skills/ | mcp.json | | OpenClaw | commands/.md | agents/.md | skills/ | TS entry point | | Qwen | commands/.md | agents/*.yaml | skills/ | N/A |

> Agentless note (V3.1): the plugin ships no standalone agents, so the Agents column is empty in practice — the converters still exist but iterate an empty set. Specialist personas travel inside skills/ (under references/personas/) and are converted as part of each skill, reaching every platform.

***

Workflow Entry Points

The nine-step compound loop, plus the full-pipeline runner and the strategy grounding skill:

| Skill | Stage | Purpose | | ---------------- | ---------- | ------------------------------------------------------------------- | | /sf-strategy | grounding | Create/maintain repo-root STRATEGY.md read by ideate/brainstorm/plan | | /sf-ideate | bread | Decide what's worth building — grounded idea generation | | /sf-brainstorm | loop | Explore requirements through collaborative dialogue | | /sf-plan | loop | Research & design specs with parallel persona research (NO CODE) | | /sf-deepen | loop | Enhance plan sections with parallel deep research | | /sf-work | loop | Implement with pre-research, skills routing, and test checks | | /sf-review | loop | Review with parallel persona dispatch (fast/thorough/comprehensive) | | /sf-polish | bread | Stack-aware UI polish — SLDS2/UX, WCAG accessibility, copy | | /sf-compound | loop | Capture learnings to docs/solutions/ with YAML schema | | /sf-lfg | pipeline | Full autonomous pipeline — ideate through deploy in one command |

Plus utility skills: /sf-simplify-code, /sf-product-pulse, /sf-debug, /sf-doc-review, /sf-optimize, /sf-resolve-pr-feedback, /sf-commit, /sf-commit-push-pr, /sf-pr-description, /sf-release-notes, /sf-report-bug, /sf-sessions, /sf-setup, and more.

/sf-lfg — The Full Pipeline

Ideate → Brainstorm → Plan → Deepen → Work → Review → Resolve → Polish → Test → Deploy → Compound

Each stage has gates that must pass before proceeding. The pipeline aborts and asks for input on security regressions, governor regressions, repeated test failures, or deployment validation problems.

/sf-lfg "Lead auto-assignment flow based on territory" --deploy=scratch

***

Specialist Personas (60)

V3.1 is agentless — there are no standalone registered agents. The 60 specialist personas are prompt assets under skills/<owner>/references/personas/<name>.md, dispatched by the workflow skills as isolated subagents: parallel with isolated context on Claude Code, applied inline on harnesses without a subagent primitive. They ship to every platform as ordinary skill files. Primary owners: sf-review (code review), sf-doc-review (doc review), sf-plan (research). Topical groupings:

| Group | Covers | | -------------------------- | ---------------------------------------------------------------------------- | | Apex | Governor limits, security (CRUD/FLS, injection), bulkification, triggers, test coverage, exceptions | | LWC | Architecture, performance, security (XSS/Locker), accessibility, Aura migration | | Flow & Automation | Flow governor limits, complexity, Flow-vs-Apex strategy, validation rules | | Integration | REST API design, callout patterns, Platform Events, integration security, MCP config & tool builder | | Architecture & Data | Data model, sharing/OWD security, pattern recognition, metadata consistency | | Research | Learnings, best practices, git history, repo conventions, framework docs | | Workflow | Spec/flow analysis, bug reproduction, PR comment resolution, simplicity, deployment verification | | Review personas | Correctness, maintainability, testing, project-standards (always-on) + conditional personas (adversarial, security, performance, reliability, API contract, data migration, …) |

***

Skills (60)

Domain Knowledge

| Skill | Scope | Use When | | ---------------------- | ---------------- | ------------------------------------- | | governor-limits | Universal | Any Apex, Flow, or trigger work | | apex-patterns | Apex only | Apex classes, triggers, services | | flow-patterns | Automation only | Building any type of Flow | | lwc-patterns | LWC only | Lightning Web Components | | graphql-patterns | LWC only | LWC GraphQL wire adapter / LDS GraphQL | | security-guide | Universal | CRUD/FLS, sharing, permissions | | integration-patterns | Integration only | Callouts, APIs, Platform Events | | test-factory | Apex only | Test classes, test data factories |

Generating Skills (from forcedotcom/afv-library, Apache-2.0)

| Skill | Generates | | ------------------------- | -------------------------------------------------------------- | | apex-generate | Apex class + tests as one unit | | flow-generate | Flow via 3-step MCP pipeline | | validation-rule-generate| Validation rules with user-friendly errors | | apex-trigger-refactor | One-trigger-per-object handler refactors | | slds2-uplift | SLDS2 styling hooks / design-token uplift | | metadata-generate | --type-dispatched object / field / app / tab / list-view / lightning-type | | lightning-page-generate | FlexiPage or full LEX app orchestration | | permission-set-generate | Least-privilege permission sets |

Agentforce & Prompt Builder

| Skill | Use When | | -------------------- | ----------------------------------------------------------------------- | | agentforce-develop | Build, modify, debug, deploy Agentforce agents — Agent Spec gate, .agent authoring, publish/activate | | agentforce-test | Smoke + batch testing — sf agent preview traces, Testing Center YAML, fix loop | | agentforce-observe | Production observation — STDM session traces in Data Cloud (with fallback) | | prompt-builder | Prompt templates — metadata XML, merge fields, grounding, deployment |

> ⚠️ Agentforce DX Critical Notes (April–May 2026): > - topic is deprecated. Use subagent, start_agent agent_router:, and @subagent.name everywhere. > - Default preview = simulated. Without --use-live-actions, real Apex is never called. --mode live does not exist — the correct flag is --use-live-actions. > - Debug logs → Agent User, not admin. Apex runs as the Einstein Agent User; setting debug logs on your admin account produces nothing. > - API version must match your org. Spring '26 = 66.0, Summer '26 = 67.0. Mismatches cause Invalid api version errors. > - Multi-component deploys need Package XML. Use --manifest manifest/package.xml, not --metadata. The metadata type is AiAuthoringBundle.

Hosted MCP

| Skill | Use When | | -------------------- | ----------------------------------------------------------------------- | | hosted-mcp-servers | Hosted MCP setup, ECA configuration, URL patterns, security model, troubleshooting | | mcp-tool-builder | Building custom MCP tools — Apex @InvocableMethod, Flows, Named Queries, prompt templates |

Tooling

| Skill | Use When | | --------------------- | ---------------------------------------------- | | sf-cli | Deploy, retrieve, test, org management | | compound-docs | Writing solution documents with YAML schema | | file-todos | File-based task tracking | | git-worktree | Isolated parallel development branches | | create-agent-skills | Creating new agents and skills for the plugin |

***

Knowledge System

Learnings are captured in docs/solutions/ with YAML frontmatter and organized by category:

docs/solutions/
├── governor-limit-issues/   # Limit handling patterns
├── deployment-issues/       # CI/CD and deployment fixes
├── test-failures/           # Test troubleshooting guides
├── security-issues/         # Security implementations
├── integration-issues/      # External system patterns
├── flow-issues/             # Automation solutions
├── lwc-issues/              # Component solutions
├── data-model-issues/       # Schema design decisions
├── best-practices/          # Proven patterns and approaches
└── patterns/                # Reusable code patterns

The sf-learnings-researcher persona searches these documents by frontmatter metadata before every plan and implementation phase to surface relevant institutional knowledge. docs/solutions/, docs/plans/, and docs/brainstorms/ are protected directories — edit, never delete.

Architecture workflow

<img width="2112" height="4080" alt="image" src="https://github.com/user-attachments/assets/afdaf91d-ef64-43e7-aa4d-c2cd711e5fbc" />

***

Project Structure

salesforce-compound-engineering-plugin/
├── .claude-plugin/
│   ├── plugin.json           # Plugin manifest (v3.1.0-beta.3)
│   └── marketplace.json      # Marketplace loader schema
├── .cursor-plugin/           # Cursor plugin manifest
├── .codex-plugin/            # Codex plugin manifest
├── .mcp.json                 # Context7 + Salesforce DX MCP config
├── schema.yaml               # YAML validation for docs/solutions/
├── PRINCIPLES.md             # Seven governing principles (source of truth)
├── CLAUDE.md                 # Project context and protected artifacts
├── cli/                      # Multi-tool installer CLI (Bun)
│   ├── package.json          # @gellasangameshgupta/sf-compound-plugin
│   ├── src/
│   │   ├── index.ts          # CLI entry (citty)
│   │   ├── parser/           # Plugin reader + markdown parser
│   │   ├── converters/       # 11 platform converters
│   │   ├── transforms/       # Path, reference, frontmatter rewriting
│   │   └── utils/            # Auto-detect, merge helpers
│   └── tests/
├── skills/                   # 60 skills — each owns the personas it dispatches
│   ├── index.md              # Skill routing map
│   ├── sf-review/references/personas/      # ~40 code-review personas
│   ├── sf-doc-review/references/personas/  #  8 doc-review personas
│   ├── sf-plan/references/personas/        #  9 research personas
│   └── …                     # 60 personas total — agentless, no standalone agents/ dir
└── docs/
    ├── brainstorms/          # Pre-planning exploration records (protected)
    ├── plans/                # Feature plans (protected)
    └── solutions/            # Institutional knowledge (protected)

***

MCP Integration

Configured in .mcp.json:

{
  "mcpServers": {
    "context7": {
      "type": "http",
      "url": "https://mcp.context7.com/mcp"
    },
    "salesforce-dx": {
      "command": "npx",
      "args": ["-y", "@salesforce/mcp", "--orgs", "DEFAULT_TARGET_ORG", "--toolsets", "all"]
    }
  }
}

Context7 — framework documentation, used by research personas as the second tier after local skills, before falling back to web search.

Salesforce DX MCP — live org operations (SOQL, deploy, retrieve, code analysis, LWC experts, testing). Toolsets: core, orgs, metadata, data, users, code-analysis, lwc-experts, aura-experts, experts-validation, devops, enrichment, mobile, testing, scale-products.

> Prerequisites for Salesforce DX MCP: Authorize an org first with sf org login web. The server uses DEFAULT_TARGET_ORG — whatever you set with sf config set target-org.

***

Requirements

• Claude Code (or another supported AI coding tool)
• Node.js (for MCP servers)
• Bun (for the CLI installer)
• Git (recommended)
• Salesforce CLI (sf) for org operations

***

Contributing

Contributions welcome! Key areas:

• Add new personas for specialized reviews
• Expand skills with more patterns
• Improve index files for better routing
• Add solution documents to docs/solutions/

See skills/create-agent-skills/SKILL.md for agent/skill authoring guidance, and read PRINCIPLES.md before non-trivial changes to workflow skills.

***

License

MIT License

***

Credits

• Built for the Salesforce developer community
• Inspired by EveryInc's Compound Engineering Plugin
• Inspired by GitHub Spec-Kit
• Generating & Agentforce skills adapted from forcedotcom/afv-library (Apache-2.0)
• Seven-principles framework distilled from Andrej Karpathy's Y Combinator AI Startup School talk
• Powered by Claude Code