Validates that file-IO tool arguments stay within the workspace root.

Parameters for a single arterial echo reflection.

Controls which tools can execute automatically vs require approval.

An echo detected from cepstral peak analysis.

Runs commands via subprocess on the host machine.

Uniform result from any executor.

Runs commands inside an E2B cloud sandbox. Requires the ``e2b_code_interpreter`` package and an ``E2B_API_KEY`` environment variable. Both a

WebSocket client that forwards tool invocations to the OpenClaw Gateway.

HTTP client that forwards tool invocations to a run_sim3d.ts --serve tool server.

Background task scheduler with cron-style and one-shot support.

Simulated robot responses for testing without hardware.

HTTP handler that translates REST calls to tool invocations.

Abstract base for shell execution backends.

Test the _parse_frontmatter function used by agent_runtime.py.

Verify agent_runtime.py has the expected pipeline methods.

Non-redirect agents in system/agents/ must be discoverable in .claude/agents/.

Validate that the regex used in show_banner() correctly extracts the art.

The bridge must return identical responses regardless of caller.

CLI argument parsing includes --strategy and --project-dir.

StepResult and StepValidation dataclasses exist and have expected fields.

Dialect_Benchmark.md is parseable with valid frontmatter for pipeline execution.

Validate each .dialect.md file.

Every dialect file must have a Grammar / Syntax section for AOT extraction.

Validate system/dialects/_index.md.

Verify all expected directories exist.

Validate base.md and index.md for the dialects skill domain.

_generate_agent_spec creates markdown agent files from scenario context.

_handle_delegate_to_agent now runs a multi-turn tool loop.

Verify that _load_manifest correctly parses GEMINI.md shell-based tools.

Verify the gemma-openrouter provider is correctly configured.

Verify the gemma provider is correctly configured in PROVIDER_CONFIGS.

Verify human-renderer-agent can detect and expand each dialect back to prose. The human-renderer defines a detection table (Content Pattern → Det

Verify intent-compiler-agent routes to correct dialect for each domain. The intent-compiler defines a mapping table (Intent Pattern → Preferred D

Verify bidirectional coverage: every dialect the intent-compiler can route TO must be detectable and expandable BY the human-renderer. This v

MODEL_CAPABILITIES lookup and fallback behavior.

Verify orchestration/index.md lists facade agents.

Pipeline parsing from scenario files (Format A and B).

Unit tests for parse_duration() — pure stdlib function.

Unit tests for parse_time() — pure stdlib function.

_parse_tool_calls extracted method handles all 4 formats.

Direct PathPolicy.validate_path() tests.

Path validation runs before mode check in authorize().

Prior step files are pre-read and injected into later step prompts.

process_input() should wrap bare goals in skillos execute: format.

QWEN.md must instruct the runtime to read Boot.md first.

QWEN.md must define all required native tools.

All tool_call examples in QWEN.md must use valid XML-like format.

run_scenario.py exists and has expected structure.

Validate requires_dialects and pipeline fields in Dialect_Benchmark.md.

Unit tests for the Scheduler class.

execute_scenario routes to the correct method based on model tier.

Validate Quick Grammar Reference and Pipeline Execution Mode in system-agent.md.

Tool-call scaffolding injects a few-shot example for mid-tier models.

_validate_step_output rule-based quality gate.