Inline workflow executor for proxy providers (ZenMux, custom API providers). Executes workflow steps directly in the main context WITHOUT spawning Task subagents. Use this skill when: - Running looplia workflows via ZenMux or other proxy providers - Task subagents fail with "invalid_model" errors - You need inline execution without context isolation Architecture: Each workflow step is executed INLINE (no Task tool) - read skill, execute mission, write output, then proceed to next step. All steps share the main context. v0.6.6: Created for cross-provider compatibility with ZenMux.
Executes looplia workflows via proxy providers (ZenMux) by running steps inline without Task subagents. Use when subagents fail with "invalid_model" errors or when the system hints for inline execution.
/plugin marketplace add memorysaver/looplia-core/plugin install looplia-core@loopliaThis skill inherits all available tools. When active, it can use any tool Claude has access to.
Execute looplia workflows without Task subagents. This skill is specifically designed for proxy providers (ZenMux, custom) where the Claude Agent SDK's subagent spawning doesn't work due to model name incompatibility.
Use this skill when:
DO NOT spawn Task subagents. Execute all workflow steps directly in the main context.
For each workflow step:
Skill("{step.skill}")Same as standard workflow-executor:
Generate sandbox ID:
{first-input-name}-{YYYY-MM-DD}-{random4chars}
Example: video-transcript-2025-12-18-xk7m
Create folder structure:
sandbox/{sandbox-id}/
├── inputs/
│ └── {input-files}.md
├── outputs/
├── logs/
└── validation.json
Copy input files to inputs/
workflows/{workflow-id}.mdskill: and mission: fieldsneeds: fieldsExecute steps ONE AT A TIME, INLINE (no Task tool):
FOR EACH step in dependency order:
│
▼
┌─────────────────────────────────────────┐
│ 1. INVOKE skill: Skill("{step.skill}") │
│ → This loads skill context │
│ │
│ 2. READ input file (if provided) │
│ → Use Read tool │
│ │
│ 3. EXECUTE mission │
│ → Follow skill instructions │
│ → Generate JSON output │
│ │
│ 4. WRITE output file │
│ → Use Write tool │
│ → Output to step.output path │
│ │
│ 5. VALIDATE output │
│ → Check required_fields │
│ → Retry if failed (max 2x) │
│ │
│ 6. UPDATE validation.json │
│ → Mark step as validated: true │
└─────────────────────────────────────────┘
│
▼
NEXT STEP
For step:
- id: analyze-content
skill: media-reviewer
mission: |
Deep analysis of video transcript. Extract key themes,
important quotes, and narrative structure.
input: ${{ sandbox }}/inputs/content.md
output: ${{ sandbox }}/outputs/analysis.json
validate:
required_fields: [contentId, headline, keyThemes]
Inline execution sequence:
Invoke skill:
Skill("media-reviewer")
Read input:
Read("sandbox/video-2025-12-18-xk7m/inputs/content.md")
Execute mission:
Write output:
Write("sandbox/video-2025-12-18-xk7m/outputs/analysis.json", jsonContent)
Validate:
Proceed to next step
❌ WRONG - Spawning Task subagents:
{
"subagent_type": "workflow-step",
"description": "Execute step...",
"prompt": "..."
}
Task subagents don't work with proxy providers.
❌ WRONG - Batching all steps: Execute all steps at once without validation between them.
✅ CORRECT - Inline step-by-step:
Same as standard workflow-executor:
| Variable | Resolution |
|---|---|
${{ sandbox }} | sandbox/{sandbox-id} |
${{ inputs.{name} }} | sandbox/{id}/inputs/{name}.md |
${{ steps.{id}.output }} | Output path of step {id} |
| Scenario | Action |
|---|---|
| Workflow not found | Error with available workflows |
Step missing skill: | Error: "Step missing required 'skill' field" |
Step missing mission: | Error: "Step missing required 'mission' field" |
| Validation fails | Retry inline with feedback (max 2 retries) |
| Max retries exceeded | Report failure with details |
| Aspect | workflow-executor | workflow-executor-inline |
|---|---|---|
| Execution | Task subagent per step | Inline in main context |
| Context | Isolated per step | Shared main context |
| Provider Support | Anthropic Direct only | All providers |
| Use Case | Production (Anthropic) | Proxy providers (ZenMux) |
Creating algorithmic art using p5.js with seeded randomness and interactive parameter exploration. Use this when users request creating art using code, generative art, algorithmic art, flow fields, or particle systems. Create original algorithmic art rather than copying existing artists' work to avoid copyright violations.
Applies Anthropic's official brand colors and typography to any sort of artifact that may benefit from having Anthropic's look-and-feel. Use it when brand colors or style guidelines, visual formatting, or company design standards apply.
Create beautiful visual art in .png and .pdf documents using design philosophy. You should use this skill when the user asks to create a poster, piece of art, design, or other static piece. Create original visual designs, never copying existing artists' work to avoid copyright violations.