Skills-First Architecture
Workflows declare skills + missions in natural language. Claude determines the best approach. No rigid agent definitions required.
A new approach to AI workflows — skills as first-class citizens, not rigid code.
Skills-First Architecture
Workflows declare skills + missions in natural language. Claude determines the best approach. No rigid agent definitions required.
Workflow-as-Markdown
Define workflows in readable YAML frontmatter + Markdown. Version control friendly, human-readable, easy to share and modify.
Build with AI
Describe what you want in plain English. looplia build generates a complete workflow definition automatically.
Sandbox Isolation
Each run creates an isolated sandbox with inputs, outputs, and logs. Pause and resume anytime with —sandbox-id.
Skills-first architecture for intelligent workflow automation
A single skill-executor handles ALL workflow steps. Every step uses the same pattern: skill: + mission:. No per-skill wrapper agents needed.
Two plugin types work together: looplia-core (infrastructure skills) and looplia-writer (domain skills). Each skill defines its own JSON output schema.
Hook-based validation tracks completion per step via validation.json. Sandbox isolation enables pause and resume with full execution context.

See how domain skills transform content into creative fuel for writers. Each step produces structured output that the next step builds upon.
Extracts themes, quotes, and key concepts from source material. Identifies patterns and relationships in the content.
→ summary.jsonNo more re-watching or re-reading to find that perfect quote
Generates creative hooks, writing angles, and thought-provoking questions. Builds on the analysis to spark ideas.
→ ideas.jsonA brainstorming partner that never runs dry
Combines analysis and ideas into an actionable writing brief with outlines, angles, and source references.
→ writing-kit.jsonYour creative brief, ready to inspire
---name: writing-kitversion: 1.1.0description: Transform content into a comprehensive writing kit
steps: # Step 1: Deep content analysis - id: summary skill: media-reviewer mission: | Deep analysis of content to extract key themes. Extract minimum 3 verbatim quotes, at least 5 key points. input: ${{ sandbox }}/inputs/content.md output: ${{ sandbox }}/outputs/summary.json
# Step 2: Creative ideation (depends on analysis) - id: ideas skill: idea-synthesis mission: Generate creative hooks, angles, and questions. needs: [summary] input: ${{ steps.summary.output }} output: ${{ sandbox }}/outputs/ideas.json
# Step 3: Assemble the writing kit (final deliverable) - id: writing-kit skill: writing-kit-assembler mission: Combine insights into an actionable writing brief. needs: [summary, ideas] input: summary: ${{ steps.summary.output }} ideas: ${{ steps.ideas.output }} output: ${{ sandbox }}/outputs/writing-kit.json final: true---Looplia handles the research grunt work so humans can focus on what they do best: crafting compelling narratives.
Choose your AI provider. Switch models with a single command.
export ANTHROPIC_API_KEY=sk-ant-...looplia run writing-kit --file article.mdlooplia config provider preset ZENMUX_ZAI_GLM47looplia config provider set auth-token sk-xxxlooplia run writing-kit --file article.md16+ Model Presets
Anthropic Claude (Haiku, Sonnet), Google Gemini, OpenAI GPT, xAI Grok, DeepSeek, and more via ZenMux.
Cost Optimization
Switch between providers based on cost, speed, or capability requirements per workflow.
From zero to first workflow in under 5 minutes.