Skip to content

execution-plans.md

Latest commit

 

History

History
25 lines (18 loc) · 1.26 KB

execution-plans.md

File metadata and controls

25 lines (18 loc) · 1.26 KB
Raw

Execution Plans (DAG)

Execution Plans allow supercli to decouple the intent to execute a command from the actual execution itself. This gives agents (and humans) the ability to dry-run commands, verify correctness, analyze side effects, and explicitly authorize actions.

Key Features

  • 4-Step Directed Acyclic Graph (DAG): Computes the precise chain of operations needed to run a command:
    1. resolve: Lookup the command spec and backend adapter configuration.
    2. validate: Run input schema validation.
    3. adapter: Prepare the external request (HTTP, MCP, OpenAPI) based on the input.
    4. transform: Process the raw response into the finalized output format.
  • Risk Assessment: Automatically classifies execution intent based on the presence of side-effects (safe for reads, medium for mutations).
  • Stateful Short-Lived Plans: The backend safely stores generated execution plans with unique IDs, locking the execution context so it cannot be tampered with between generation and execution.

Usage

# 1. Generate an execution plan without running the command
supercli plan <namespace> <resource> <action> --arg1 value

# (Output returns a plan_id and a list of steps)

# 2. Execute the previously created plan
supercli execute <plan_id>