By ialameh
Persistent project memory + local-LLM offload + Salesforce domain skills for Claude Code
add-feature workflow ā see skills/add-feature/SKILL.md or skills/coding/add-feature/SKILL.md or skills/reasoning/add-feature/SKILL.md
Full agentic multi-file refactor ā planner ā coder ā qa-reviewer ā qa-fixer pipeline w/ rollback. See skills/workflow/agent/SKILL.md
analyze workflow ā see corresponding skills/*/analyze/SKILL.md
Scaffold FFLib enterprise patterns.
api workflow ā see corresponding skills/*/api/SKILL.md
Use for analysis of NON-CODE sources ā specs, PRDs, user research, support tickets, sales transcripts, log dumps, CSV/JSON datasets. Read-only. Extracts insight, finds patterns, surfaces contradictions. Sibling to investigator (which targets code).
Use to refactor Apex classes/triggers for bulk safety. Targeted single-purpose pass ā finds SOQL/DML in loops, refactors to bulk-safe collections, preserves behaviour. Returns a diff and a justification.
Use for cross-codebase integration analysis ā two repos that need to interoperate. Maps gaps, generates integration spec, recommends bridge architecture. Read-only on both sides. Pairs with /siftcoder:bridge command.
Use to implement an approved plan. Refuses scope creep. Strict adherence to the plan; surfaces deviations for user approval before deviating.
Use to generate documentation in a specific shape ā architecture, API ref, user manual, technical, runbook, migration. Pairs with the `document` skill. Cites every claim.
Use when adding a new capability to an existing project. Smaller than build (no full spec) but bigger than fix. Incremental, behind a feature flag where appropriate, with rollback path.
Use to auto-resolve build errors. Reads compiler/linker/bundler output, classifies, suggests minimal-diff fixes. Different from /heal (which is multi-strategy retry across build/test/lint).
Use when the user says "build X", "implement X", "create X feature" or hands a spec and expects working code. Spec-first, evidence-gated, test-as-acceptance. Refuses to start coding before alignment on what success looks like.
Use for generic debugging ā error analysis, stack trace parsing, issue reproduction, git bisect, code tracing. Different from /sf-debug (Salesforce-specific). Pairs with /investigate skill (read-only diagnosis).
Use when the user reports a bug, says "fix X", "X is broken", "X doesn't work". Bounded-scope, root-cause-first. Adds a regression test before declaring done. Refuses to expand into refactor.
Admin access level
Server config contains admin-level keywords
Executes bash commands
Hook triggers when Bash tool is used
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.
Modifies files
Hook triggers on file write and edit operations
Modifies files
Hook triggers on file write and edit operations
Uses power tools
Uses Bash, Write, or Edit tools
Uses power tools
Uses Bash, Write, or Edit tools
Persistent project memory + local-LLM offload + Salesforce domain skills for Claude Code.
š Full documentation: ialameh.github.io/sift-coder ā quickstart, mental model, architecture, every CLI command, every skill, every agent, troubleshooting, and a cookbook of real workflows.
SiftCoder is a Claude Code plugin that adds four things Claude Code does not have natively:
mem_why), decay, and RRF retrieval.Claude Code already has Plan/Explore/general-purpose subagents, native skills, native MCP, hooks, plan mode, /review, /security-review, prompt caching, and tool-use orchestration. What it lacks:
Requires Node ā„ 20.
git clone https://github.com/ialameh/sift-coder.git
cd sift-coder
npm install
npm run build
In Claude Code, register the plugin:
# from any project
/plugin install /path/to/siftcoder
Recommended: install Ollama for local-LLM benefits.
brew install ollama && brew services start ollama
ollama pull nomic-embed-text llama3.2:3b
node bin/siftcoder.mjs setup
# spawn the memory daemon for your workspace
node bin/siftcoder.mjs start
# verify
node bin/siftcoder.mjs status
Inside Claude Code:
/siftcoder:mem status
/siftcoder:mem drain
/siftcoder:ideate
/siftcoder:sf-deploy validate
/siftcoder:reverse-prompt deep
Each skill is a workflow contract ā explicit ordering, evidence gates, output shape, anti-patterns. Auto-loaded by trigger description; also invokable via slash command.
| Family | Skills |
|---|---|
| coding (14) | build, fix, add-feature, investigate, heal, tdd, pair, refactor, optimize, zen, debug, build-fix, perf, test |
| reasoning (7) | archaeologist, oracle, ghost, duck, dream, fortune, narrator |
| quality (9) | chaos, fuzz-mind, invariant, empathy, ripple, timewarp, blast-radius, polyglot, comply |
| review (2) | review, security |
| knowledge (5) | memory-usage, pattern-learn, pattern-search, knowledge, search |
| docs (8) | document, update-docs, update-codemaps, codemap, codemap-fast, codemap-diff, codemap-trust, codemap-export |
| workflow (14) | autonomous, swarm, pause, continue, handoff, checkpoint, session-eval, smart-retry, preview, scope, chroot, agent, focus, organize-project |
| spec (5) | improve-spec, spec-from-stories, gap-analysis, reverse-spec, feasibility |
| salesforce (12) | salesforce-apex, salesforce-lwc, salesforce-deploy, salesforce-architecture, salesforce-test, salesforce-agentforce, salesforce-einstein, salesforce-cpq, salesforce-comply, salesforce-flow, salesforce-security, schema-migrate |
| integrations (4) | api, integrate, migrate, bridge |
| ux (3) | ideate, surprise-me, reverse-prompt |
| meta (13) | siftcoder, onboard, prompt, team, sync, monitor, trace, analyze, compression, local-llm-setup, quality-check, budget, wizard |
Each skills/<family>/<name>/SKILL.md body documents: method, output shape, hard rules, anti-patterns, when-NOT-to-use, and value-add over native Claude Code.
Claude Code bridge for SiftMemory resumable reasoning memory.
npx claudepluginhub ialameh/sift-coder --plugin siftcoderMemory compression system for Claude Code - persist context across sessions
Multi-model consensus engine integrating OpenAI Codex CLI, Gemini CLI, and Claude CLI for collaborative code review and problem-solving.
Standalone image generation plugin using Nano Banana MCP server. Generates and edits images, icons, diagrams, patterns, and visual assets via Gemini image models. No Gemini CLI dependency required.
Ultra-compressed communication mode. Cuts ~75% of tokens while keeping full technical accuracy by speaking like a caveman.
Comprehensive UI/UX design plugin for mobile (iOS, Android, React Native) and web applications with design systems, accessibility, and modern patterns
Write feature specs, plan roadmaps, and synthesize user research faster. Keep stakeholders updated and stay ahead of the competitive landscape.