HAL9000
18d00c04c4
CI / lint (pull_request) Failing after 1m15s
CI / quality (pull_request) Successful in 1m21s
CI / typecheck (pull_request) Successful in 1m34s
CI / security (pull_request) Successful in 1m37s
CI / coverage (pull_request) Has been skipped
CI / unit_tests (pull_request) Failing after 1m37s
CI / docker (pull_request) Has been skipped
CI / build (pull_request) Successful in 33s
CI / helm (pull_request) Successful in 26s
CI / push-validation (pull_request) Successful in 19s
CI / e2e_tests (pull_request) Successful in 3m20s
CI / integration_tests (pull_request) Successful in 4m32s
CI / status-check (pull_request) Failing after 3s
Implements AgentSkillDiscovery class to support discovering Agent Skills from multiple configured directories across three scopes (global, project, local). Handles name collisions with precedence: local > project > global. Adds comprehensive BDD test coverage for multi-scope discovery scenarios including: - Global-only, project-only, and local-only discovery - Combined discovery from all scopes - Name collision resolution with proper precedence - Non-existent and empty scope directory handling - Multiple skills in same scope discovery ISSUES CLOSED: #9369
3.2 KiB
3.2 KiB
Tool & Validation Domain Models
The Tool and Validation domain models are defined in
src/cleveragents/domain/models/core/tool.py.
Overview
| Model | Purpose |
|---|---|
Tool |
Atomic unit of execution — namespaced, independently registered callable operation |
Validation |
Extends Tool with pass/fail semantics — always read-only, never writes, never checkpointable |
Key classes
ToolSource— enum:mcp,agent_skill,builtin,custom,wrappedToolType— enum:tool,validationValidationMode— enum:required,informationalToolCapability— capability metadata (read_only, writes, checkpointable, side_effects, etc.)ResourceSlot— typed resource binding declarationToolLifecycle— optional lifecycle hooks (discover, activate, deactivate)
Factory methods
Tool.from_config(config)— create a Tool from a YAML configuration dictValidation.from_config(config)— create a Validation from a YAML configuration dict
Serialization
tool.as_cli_dict()— stable-ordered dictionary for CLI rendering
Testing
Behave BDD tests (unit tests)
The primary test suite lives in features/tool_model.feature with step
definitions in features/steps/tool_model_steps.py.
Run with nox:
nox -s unit_tests -- features/tool_model.feature
The Behave suite covers:
- Tool/Validation model creation and field validation
- Source-conditional field requirements
- Capability constraint enforcement
- ResourceSlot binding validation
from_config()loading (including YAML example files)as_cli_dict()output stability- Enum value completeness
Robot Framework smoke tests
The Robot smoke suite in robot/tool_model.robot provides integration-level
smoke tests that verify the same model creation and YAML loader outputs
through a subprocess helper (robot/helper_tool_model.py).
Run with nox:
nox -s integration_tests -- robot/tool_model.robot
Or directly with Robot Framework:
robot --outputdir build/reports/robot robot/tool_model.robot
The Robot suite covers:
| Test Case | What it verifies |
|---|---|
| Create Minimal Tool Model | Builtin tool creation, identity fields |
| Create Validation Model With Forced Constraints | Validation read_only/writes/checkpointable forcing |
| Load Custom Tool From YAML Config | Tool.from_config() with examples/tools/custom-tool.yaml |
| Load MCP Tool From YAML Config | Tool.from_config() with examples/tools/mcp-tool.yaml |
| Load Wrapped Validation From YAML Config | Validation.from_config() with examples/validations/wrapped-validation.yaml |
| Load Required Validation From YAML Config | Validation.from_config() with examples/validations/required-validation.yaml |
| Validate Forced Constraints On Writable Capability | Validation forces read_only even with writable capability |
| Reject Invalid Tool Configurations | Invalid name, missing code, conflicting capabilities rejected |
ASV benchmarks
Performance benchmarks live in benchmarks/tool_model_bench.py:
nox -s benchmark