Creating Pipelines

What's a Pipeline?

A pipeline defines a multi-agent workflow — an ordered sequence of agent steps where each step's output feeds into the next. Pipelines support branching, loops, and resumability.

Pipeline Format

name: my-pipeline
version: 1.0.0
description: "What this pipeline accomplishes"
trigger: "natural language pattern *"

steps:
  - name: step-one
    agent: first-agent
    input: "what this step receives"
    output: "what this step produces"
    on-failure: halt

  - name: step-two
    agent: second-agent
    input: "output from step:step-one"
    output: "final result"
    on-failure: loop
    loop-target: step-one
    max-iterations: 3

resumable: true

The Complexity Router Pre-Step

Before any pipeline executes, the complexity-router automatically runs as a pre-step:

1. Classification — Analyzes the request complexity
2. Model Selection — Chooses the optimal model tier
3. Pipeline Selection — Determines if this is the right pipeline or if a simpler/more complex one should be used

This routing happens transparently. The router uses signals from: - Task scope and dependencies - Required domain expertise - Expected output complexity - Time constraints

See skills/complexity-router/resources/complexity-signals.md for the full classification criteria.

Step Configuration

on-failure options

• halt — Stop the pipeline; report failure
• loop — Go back to loop-target step (requires max-iterations)
• skip — Skip this step; continue to next
• retry — Retry this step (uses max-iterations)

Input References

Use step:name to reference output from a previous step:

• "Spec from step:specify" — uses the specify step's output
• "Code from step:implement AND spec from step:specify" — multiple inputs

Existing Pipelines

See the pipelines directory for all available pipelines.

v2.0: Pipeline Execution Engine

In v2.0, pipelines are actually executed by the PipelineExecutor engine. This replaces the v0.x approach where pipelines were just YAML documentation of intended workflows.

Running a Pipeline

# Via CLI
omniskill pipeline run sdd-pipeline --project ./myapp

# Via SDK
from sdk.omniskill import OmniSkill
os = OmniSkill()
os.execute_pipeline("sdd-pipeline", project_name="myapp")

Pipeline State Machine

The engine manages pipeline state through these transitions:

pending → validating → executing → completed
                         ↓
                    failed / paused / cancelled

• pending — Pipeline loaded, not yet started
• validating — Checking prerequisites and schema compliance
• executing — Running steps sequentially
• completed — All steps finished successfully
• failed — A step failed with on-failure: halt
• paused — Execution paused (can be resumed)
• cancelled — Manually cancelled

State Persistence

Pipeline state is saved as JSON to ~/.copilot/.omniskill/pipeline-states/. This enables:

• Resumability — omniskill pipeline resume <state-id>
• Audit trail — Every step, decision, and deviation is recorded
• Accumulated state — Decisions, constraints, and tech_stack grow across steps (never shrink)

Context Curation Between Steps

The context-curator agent runs between pipeline steps to create focused context briefs:

1. Summarizes the previous step's output
2. Filters to only what the next agent needs
3. Applies token budget constraints
4. Generates a structured context brief

This prevents context pollution — each agent gets a clean, focused input rather than the entire accumulated history.

Artifact Validation

The ArtifactValidator checks outputs between steps:

• validate_exists(path) — File must exist
• validate_sections(path, headings) — Required Markdown sections
• validate_min_content(path, words) — Minimum content length
• validate_schema(path, schema) — JSON/YAML schema compliance
• validate_compliance_score(report, threshold) — Score gate

CLI Commands

omniskill pipeline run <name> --project <dir>   # Execute a pipeline
omniskill pipeline status [state-id]             # Show current status
omniskill pipeline resume <state-id>             # Resume paused pipeline
omniskill pipeline list                          # List all active pipelines
omniskill pipeline cancel <state-id>             # Cancel with cleanup