agentcli blueprint

Lint and validate blueprint YAML files against the platform’s Pydantic schemas. Auto-detects whether the file is an agent, strategy, or workflow blueprint and runs the appropriate validation.

Synopsis

agentcli blueprint <subcommand> [OPTIONS] YAML_PATH

Subcommands

Subcommand Description
lint Validate a blueprint YAML file and display a coverage report
validate Alias for lint

agentcli blueprint lint

agentcli blueprint lint YAML_PATH

Validates the file against the appropriate Pydantic model (AgentBlueprint, StrategyBlueprint, or WorkflowBlueprint), then runs cross-block validation to catch configuration inconsistencies. Exits with code 1 on failure, 0 on success.

Arguments

Argument Description
YAML_PATH Path to the blueprint YAML file to validate

Output

For agent blueprints, lint prints a block coverage tree showing which subsystems are configured:

my-agent.yaml (agent blueprint)
+-- Name: my-agent
+-- Version: 1.0.0
+-- Model
|   +-- provider: bedrock
|   +-- model_id: eu.anthropic.claude-sonnet-4-6
+-- Block Coverage
    +-- runtime
    +-- gateway
    +-- identity  —
    +-- memory
    +-- tools  —
    +-- observability
    +-- evaluation  —
    +-- policy  —
    +-- multi_agent  —

PASS — Blueprint is valid.

A check mark indicates the block is configured; a dash indicates it is absent (not an error).

Cross-Block Warnings

After schema validation, lint checks for configuration inconsistencies:

Warning Cause
Policy rules declared but no gateway configured policy.rules defined but no gateway block
Memory strategies require runtime.type: agentcore Memory strategies set but runtime type is not agentcore
Coordinator role declared but no multi_agent.nodes defined Coordinator pattern declared without nodes
Online evaluation configured but no observability block evaluation.online enabled without observability (evaluation reads OTEL traces)

Examples

Validate an agent blueprint

agentcli blueprint lint agents/my-agent.yaml

Validate a strategy blueprint

agentcli blueprint lint strategies/extraction-strategy.yaml

Strategy blueprints are auto-detected by the presence of required_signals, entry_conditions, or exit_conditions fields.

Validate a workflow blueprint

agentcli blueprint lint workflows/document-pipeline.yaml

Workflow blueprints are auto-detected by the presence of a states field.

Use in CI

agentcli blueprint lint agents/my-agent.yaml && echo "Blueprint OK"
# Exit code 1 on validation failure — safe to use in CI pipelines

Validate all blueprints in a directory

for f in agents/*.yaml; do agentcli blueprint lint "$f"; done

Blueprint Types and Required Fields

Agent Blueprint

Required fields: id, name, version, model (with model_id, temperature, and max_tokens), and prompt_ref. All other blocks (runtime, gateway, memory, etc.) have defaults and are optional.

id: my-agent
name: My Agent
version: 1.0.0
prompt_ref: my-agent-system          # required; Prompt Registry key
model:
  provider: bedrock                  # optional — defaults to bedrock
  model_id: ${MODEL_ID}             # required; supports ${VAR:-default} expansion
  temperature: 0.3                   # required
  max_tokens: 4096                   # required

Strategy Blueprint

Required fields: id, name, version. All other fields (required_agents, required_signals, entry_conditions, exit_conditions, etc.) are optional. Detected by the presence of required_signals, entry_conditions, or exit_conditions.

Workflow Blueprint

Required fields: id, name, version. The states field defaults to an empty list and is optional in the schema — though a workflow without states is a no-op. Detected by the presence of a states field.

See Also

  • Agent Blueprint — full YAML schema reference including all optional blocks
  • Strategy Blueprint — strategy-specific fields (required_agents, entry_conditions, exit_conditions, risk_controls)
  • Workflow Blueprint — Step Functions state machine schema (states, trigger, timeout_minutes)
  • agentcli deploy — deploy a validated blueprint to AgentCore Runtime
  • agentcli graph — visualize the multi_agent block of an agent blueprint