claude-dev-env 1.17.2 → 1.19.0

package/hooks/blocking/test_prompt_workflow_gate_core.py

@@ -1,195 +0,0 @@
-"""Unit tests for shared prompt workflow gate logic."""
-
-from prompt_workflow_gate_core import (
-    extract_fenced_xml_content,
-    extract_fenced_xml_content_from_export,
-    find_ambiguous_scope_terms,
-    has_checklist_container,
-    has_internal_object_leak,
-    is_prompt_workflow_response,
-    missing_context_control_signals,
-    missing_checklist_rows,
-    missing_required_xml_sections,
-    missing_scope_anchors,
-    normalize_prompt_workflow_export,
-)
-
-def test_internal_object_leak_detected() -> None:
-    text = '{"pipeline_mode": "internal_section_refinement_with_final_audit"}'
-    assert has_internal_object_leak(text)
-
-def test_missing_scope_anchors_returns_expected_rows() -> None:
-    text = "target_local_roots only."
-    missing = missing_scope_anchors(text)
-    assert "target_canonical_roots" in missing
-    assert "completion_boundary" in missing
-
-def test_missing_checklist_rows_detected() -> None:
-    text = "checklist_results: structured_scoped_instructions only"
-    missing = missing_checklist_rows(text)
-    assert "completion_boundary_measurable" in missing
-
-def test_checklist_container_detection() -> None:
-    assert has_checklist_container("checklist_results:\n- structured_scoped_instructions")
-
-def test_prompt_workflow_response_detection() -> None:
-    message = (
-        "overall_status: pass\n"
-        "target_local_roots: /repo\n"
-        "comparison_basis: current behavior vs deterministic guarantees\n"
-    )
-    assert is_prompt_workflow_response(message)
-
-def test_missing_context_control_signals_detected() -> None:
-    missing = missing_context_control_signals("base_minimal_instruction_layer: true")
-    assert "on_demand_skill_loading: true" in missing
-
-def test_ambiguous_scope_terms_detected() -> None:
-    text = "Scope applies to this session and current files."
-    terms = find_ambiguous_scope_terms(text)
-    assert "this session" in terms
-    assert "current files" in terms
-
-def _fenced_xml(body: str) -> str:
-    return f"```xml\n{body}\n```"
-
-def _runtime_context_lines() -> tuple[str, ...]:
-    return (
-        "<runtime_context>",
-        "base_minimal_instruction_layer: true",
-        "on_demand_skill_loading: true",
-        "</runtime_context>",
-        "",
-    )
-
-def _flattened_transcript(*lines: str) -> str:
-    return "\n".join(lines) + "\n"
-
-def _flattened_attempt(*body_lines: str, audit_line: str = "Audit: pass 15/15") -> str:
-    flattened_lines = [audit_line, ""]
-    for line in body_lines:
-        flattened_lines.append(f"  {line}" if line else "")
-    return "\n".join(flattened_lines)
-
-def test_missing_required_xml_sections_all_present_returns_empty() -> None:
-    body = (
-        "<role>R.</role>\n"
-        "<background>C.</background>\n"
-        "<instructions>I.</instructions>\n"
-        "<constraints>Co.</constraints>\n"
-        "<output_format>O.</output_format>\n"
-    )
-    assert missing_required_xml_sections(_fenced_xml(body)) == []
-
-def test_missing_required_xml_sections_missing_background() -> None:
-    body = (
-        "<role>R.</role>\n"
-        "<instructions>I.</instructions>\n"
-        "<constraints>Co.</constraints>\n"
-        "<output_format>O.</output_format>\n"
-    )
-    assert missing_required_xml_sections(_fenced_xml(body)) == ["background"]
-
-def test_missing_required_xml_sections_missing_role_and_output_format() -> None:
-    body = (
-        "<background>C.</background>\n"
-        "<instructions>I.</instructions>\n"
-        "<constraints>Co.</constraints>\n"
-    )
-    missing = missing_required_xml_sections(_fenced_xml(body))
-    assert missing == ["role", "output_format"]
-
-def test_missing_required_xml_sections_no_fence_returns_empty() -> None:
-    assert missing_required_xml_sections("no fenced xml here") == []
-
-def test_missing_required_xml_sections_prose_without_tags_counts_as_missing() -> None:
-    body = (
-        "<role>R.</role>\n"
-        "background appears in prose but has no tags.\n"
-        "<instructions>I.</instructions>\n"
-        "<constraints>Co.</constraints>\n"
-        "<output_format>O.</output_format>\n"
-    )
-    assert missing_required_xml_sections(_fenced_xml(body)) == ["background"]
-
-def test_extract_fenced_xml_preserves_content_after_nested_inner_fence() -> None:
-    message = (
-        "```xml\n"
-        "<role>R</role>\n"
-        "<illustrations>\n"
-        "```bash\necho hi\n```\n"
-        "</illustrations>\n"
-        "<background>B</background>\n"
-        "<instructions>I</instructions>\n"
-        "<constraints>C</constraints>\n"
-        "<output_format>O</output_format>\n"
-        "```\n"
-    )
-    extracted = extract_fenced_xml_content(message)
-    assert "</illustrations>" in extracted
-    assert "<background>B</background>" in extracted
-
-def test_normalize_prompt_workflow_export_rebuilds_fence_from_flattened_transcript() -> None:
-    transcript = _flattened_transcript(
-        _flattened_attempt(
-            *_runtime_context_lines(),
-            "<role>R</role>",
-            "<background>B</background>",
-            "<instructions>I</instructions>",
-            "<constraints>C</constraints>",
-            "<output_format>O</output_format>",
-            "✻ Worked for 1m 7s",
-            audit_line="● Audit: pass 15/15",
-        ),
-    )
-    normalized = normalize_prompt_workflow_export(transcript)
-    assert normalized.startswith("Audit: pass 15/15\n```xml\n")
-    assert normalized.endswith("\n```")
-    assert "<runtime_context>" in normalized
-    assert "✻ Worked for 1m 7s" not in normalized
-
-def test_normalize_prompt_workflow_export_uses_last_audit_attempt() -> None:
-    first_attempt = _flattened_attempt(
-        "<role>FIRST</role>",
-        "<background>Old</background>",
-        "<instructions>Old</instructions>",
-        "<constraints>Old</constraints>",
-        "<output_format>Old</output_format>",
-        audit_line="● Audit: pass 15/15",
-    )
-    second_attempt = _flattened_attempt(
-        *_runtime_context_lines(),
-        "<role>FINAL</role>",
-        "<background>Fresh</background>",
-        "<instructions>I</instructions>",
-        "<constraints>C</constraints>",
-        "<output_format>O</output_format>",
-        "✻ Worked for 2m 8s",
-    )
-    transcript = _flattened_transcript(
-        first_attempt,
-        "",
-        "● Re-emitting the full artifact with the runtime signals added.",
-        "",
-        second_attempt,
-    )
-    normalized = normalize_prompt_workflow_export(transcript)
-    assert "<role>FINAL</role>" in normalized
-    assert "<role>FIRST</role>" not in normalized
-
-def test_extract_fenced_xml_content_from_export_supports_flattened_transcript() -> None:
-    transcript = _flattened_transcript(
-        _flattened_attempt(
-            "<role>R</role>",
-            "<background>B</background>",
-            "<instructions>I</instructions>",
-            "<constraints>C</constraints>",
-            "<output_format>O</output_format>",
-            "✻ Worked for 31s",
-            audit_line="● Audit: pass 15/15",
-        ),
-    )
-    extracted = extract_fenced_xml_content_from_export(transcript)
-    assert extracted.startswith("<role>R</role>")
-    assert "<output_format>O</output_format>" in extracted
-    assert "Worked for" not in extracted

package/hooks/blocking/test_prompt_workflow_validate.py

@@ -1,339 +0,0 @@
-"""Tests for prompt_workflow_validate module (shared validator + CLI entry point)."""
-
-import subprocess
-import sys
-from pathlib import Path
-
-import pytest
-
-from prompt_workflow_validate import ValidationResult, validate_prompt_workflow
-
-VALIDATOR_MODULE_PATH = Path(__file__).parent / "prompt_workflow_validate.py"
-
-
-def _full_checklist_rows() -> str:
-    return (
-        "checklist_results:\n"
-        "- structured_scoped_instructions\n"
-        "- sequential_steps_present\n"
-        "- positive_framing\n"
-        "- acceptance_criteria_defined\n"
-        "- safety_reversibility_language\n"
-        "- reversible_action_and_safety_check_guidance\n"
-        "- concrete_output_contract\n"
-        "- scope_boundary_present\n"
-        "- explicit_scope_anchors_present\n"
-        "- all_instructions_artifact_bound\n"
-        "- scope_terms_explicit_and_anchored\n"
-        "- completion_boundary_measurable\n"
-        "- citation_grounding_policy_present\n"
-        "- source_priority_rules_present\n"
-        "- artifact_language_confidence\n"
-    )
-
-
-def _wrap_five_section_scaffold(inner_body: str) -> str:
-    has_instructions = "<instructions>" in inner_body
-    has_constraints = "<constraints>" in inner_body
-    instructions_section = (
-        "" if has_instructions else "<instructions>Test instructions sentence one.</instructions>\n"
-    )
-    constraints_section = (
-        "" if has_constraints else "<constraints>Test constraints sentence one.</constraints>\n"
-    )
-    return (
-        "<role>Test role sentence one.</role>\n"
-        "<background>Test background sentence one.</background>\n"
-        f"{instructions_section}"
-        f"{inner_body}\n"
-        f"{constraints_section}"
-        "<output_format>Test output format sentence one.</output_format>\n"
-    )
-
-
-def _build_prompt_workflow_message_with_fenced_xml(fenced_xml_body: str) -> str:
-    return (
-        "Audit: pass 15/15\n"
-        "```xml\n" + fenced_xml_body + "\n```\n"
-        "overall_status: pass\n" + _full_checklist_rows() + "target_local_roots\n"
-        "target_canonical_roots\n"
-        "target_file_globs\n"
-        "comparison_basis\n"
-        "completion_boundary\n"
-        "base_minimal_instruction_layer: true\n"
-        "on_demand_skill_loading: true\n"
-    )
-
-
-class TestValidatePromptWorkflowFunction:
-    """Tests that exercise the shared validate_prompt_workflow function directly."""
-
-    def test_allowed_complete_message_with_fenced_xml(self) -> None:
-        fenced_content = _wrap_five_section_scaffold(
-            "<instructions>Ensure all functions have explicit return types.</instructions>"
-        )
-        message = _build_prompt_workflow_message_with_fenced_xml(fenced_content)
-        validation_result = validate_prompt_workflow(message)
-        assert validation_result.allowed is True
-        assert validation_result.reasons == ()
-
-    def test_blocked_missing_context_control_lines(self) -> None:
-        message = (
-            "overall_status: pass\n"
-            + _full_checklist_rows()
-            + "target_local_roots\n"
-            + "target_canonical_roots\n"
-            + "target_file_globs\n"
-            + "comparison_basis\n"
-            + "completion_boundary\n"
-        )
-        validation_result = validate_prompt_workflow(message)
-        assert validation_result.allowed is False
-        assert "missing_context_signals" in validation_result.reason_codes
-        assert any(
-            "context-control" in each_message
-            for each_message in validation_result.reason_messages
-        )
-
-    def test_allowed_empty_message(self) -> None:
-        validation_result = validate_prompt_workflow("")
-        assert validation_result.allowed is True
-
-    def test_allowed_non_workflow_message(self) -> None:
-        validation_result = validate_prompt_workflow("Just a regular response.")
-        assert validation_result.allowed is True
-
-    def test_blocked_internal_object_leak(self) -> None:
-        leak_message = (
-            '{"pipeline_mode": "internal_section_refinement_with_final_audit"}'
-        )
-        validation_result = validate_prompt_workflow(leak_message)
-        assert validation_result.allowed is False
-        assert "internal_object_leak" in validation_result.reason_codes
-
-    def test_allowed_internal_object_with_debug_context(self) -> None:
-        leak_message = (
-            '{"pipeline_mode": "internal_section_refinement_with_final_audit"}'
-        )
-        validation_result = validate_prompt_workflow(
-            leak_message,
-            user_context="debug: show internal pipeline object",
-        )
-        assert validation_result.allowed is True
-
-    def test_blocked_missing_checklist_rows(self) -> None:
-        message = (
-            "overall_status: pass\n"
-            "checklist_results: structured_scoped_instructions\n"
-            "target_local_roots\n"
-            "target_canonical_roots\n"
-            "target_file_globs\n"
-            "comparison_basis\n"
-            "completion_boundary\n"
-        )
-        validation_result = validate_prompt_workflow(message)
-        assert validation_result.allowed is False
-        assert "missing_checklist_rows" in validation_result.reason_codes
-
-    def test_blocked_negative_keywords_in_fenced_xml(self) -> None:
-        fenced_content = _wrap_five_section_scaffold(
-            "<instructions>Do not leave return types implicit.</instructions>"
-        )
-        message = _build_prompt_workflow_message_with_fenced_xml(fenced_content)
-        validation_result = validate_prompt_workflow(message)
-        assert validation_result.allowed is False
-        assert "negative_keywords_in_artifact" in validation_result.reason_codes
-
-    def test_blocked_ambiguous_scope(self) -> None:
-        message = (
-            "overall_status: pass\n"
-            + _full_checklist_rows()
-            + "scope block includes target_local_roots target_canonical_roots "
-            + "target_file_globs comparison_basis completion_boundary "
-            + "base_minimal_instruction_layer: true\n"
-            + "on_demand_skill_loading: true\n"
-            + "and applies to this session."
-        )
-        validation_result = validate_prompt_workflow(message)
-        assert validation_result.allowed is False
-        assert "ambiguous_scope" in validation_result.reason_codes
-
-    def test_reason_messages_property(self) -> None:
-        message = (
-            "overall_status: pass\n"
-            + _full_checklist_rows()
-            + "target_local_roots\n"
-            + "target_canonical_roots\n"
-            + "target_file_globs\n"
-            + "comparison_basis\n"
-            + "completion_boundary\n"
-        )
-        validation_result = validate_prompt_workflow(message)
-        assert len(validation_result.reason_messages) == 1
-        assert len(validation_result.reason_codes) == 1
-
-    def test_blocked_missing_scope_anchors(self) -> None:
-        message = (
-            "overall_status: pass\n"
-            + _full_checklist_rows()
-            + "base_minimal_instruction_layer: true\n"
-            + "on_demand_skill_loading: true\n"
-        )
-        validation_result = validate_prompt_workflow(message)
-        assert validation_result.allowed is False
-        assert "missing_scope_anchors" in validation_result.reason_codes
-
-    def test_blocked_missing_xml_sections_in_fenced_artifact(self) -> None:
-        fenced_body = (
-            "<role>Test role sentence one.</role>\n"
-            "<instructions>Test instructions sentence one.</instructions>\n"
-            "<constraints>Test constraints sentence one.</constraints>\n"
-            "<output_format>Test output format sentence one.</output_format>\n"
-        )
-        message = _build_prompt_workflow_message_with_fenced_xml(fenced_body)
-        validation_result = validate_prompt_workflow(message)
-        assert validation_result.allowed is False
-        assert "missing_xml_sections" in validation_result.reason_codes
-        assert any(
-            "background" in each_message
-            for each_message in validation_result.reason_messages
-        )
-
-    def test_allows_positive_phrasing_inside_fenced_xml(self) -> None:
-        fenced_content = _wrap_five_section_scaffold(
-            "<instructions>Ensure all functions have explicit return types.</instructions>"
-        )
-        message = _build_prompt_workflow_message_with_fenced_xml(fenced_content)
-        validation_result = validate_prompt_workflow(message)
-        assert validation_result.allowed is True
-
-    def test_permits_negative_keywords_outside_fenced_xml(self) -> None:
-        fenced_inner = _wrap_five_section_scaffold(
-            "<instructions>Ensure all functions have explicit return types.</instructions>"
-        )
-        message = (
-            "Audit: pass 15/15\n"
-            "Do not skip the audit line.\n"
-            "```xml\n" + fenced_inner + "\n```\n"
-            "overall_status: pass\n" + _full_checklist_rows() + "target_local_roots\n"
-            "target_canonical_roots\n"
-            "target_file_globs\n"
-            "comparison_basis\n"
-            "completion_boundary\n"
-            "base_minimal_instruction_layer: true\n"
-            "on_demand_skill_loading: true\n"
-        )
-        validation_result = validate_prompt_workflow(message)
-        assert validation_result.allowed is True
-
-
-@pytest.mark.parametrize(
-    ("banned_pattern_name", "fenced_xml_content"),
-    [
-        ("do_not", "<instructions>Do not leave return types implicit.</instructions>"),
-        ("avoid", "<instructions>Avoid missing return types.</instructions>"),
-        ("never", "<constraints>Never store credentials in plain text.</constraints>"),
-        ("without", "<instructions>Deploy without running tests first.</instructions>"),
-        ("prevent", "<constraints>Prevent unauthorized access to the API.</constraints>"),
-        ("reject", "<constraints>Reject all unsigned commits.</constraints>"),
-        ("cannot", "<constraints>The API cannot accept unauthenticated requests.</constraints>"),
-        ("unless", "<constraints>Skip the build step unless the user explicitly approves.</constraints>"),
-        ("must_not", "<constraints>The script must not produce duplicates.</constraints>"),
-        ("must_never", "<constraints>You must never store credentials in environment variables.</constraints>"),
-        ("instead_of", "<instructions>Use explicit types instead of implicit ones.</instructions>"),
-        ("rather_than", "<constraints>Prefer explicit types rather than inferred ones.</constraints>"),
-        ("as_opposed_to", "<instructions>Use Grid as opposed to floats for layout.</instructions>"),
-    ],
-)
-def test_blocks_banned_pattern_inside_fenced_xml(
-    banned_pattern_name: str,
-    fenced_xml_content: str,
-) -> None:
-    message = _build_prompt_workflow_message_with_fenced_xml(
-        _wrap_five_section_scaffold(fenced_xml_content)
-    )
-    validation_result = validate_prompt_workflow(message)
-    assert validation_result.allowed is False
-    assert "negative_keywords_in_artifact" in validation_result.reason_codes
-
-
-class TestValidatorCli:
-    """Tests that exercise the CLI entry point via subprocess."""
-
-    def test_cli_exits_zero_for_valid_content(self, tmp_path: Path) -> None:
-        fenced_content = _wrap_five_section_scaffold(
-            "<instructions>Ensure all functions have explicit return types.</instructions>"
-        )
-        draft_file = tmp_path / "draft.xml"
-        draft_file.write_text(
-            _build_prompt_workflow_message_with_fenced_xml(fenced_content),
-            encoding="utf-8",
-        )
-        completed_process = subprocess.run(
-            [sys.executable, str(VALIDATOR_MODULE_PATH), str(draft_file)],
-            capture_output=True,
-            text=True,
-            check=False,
-        )
-        assert completed_process.returncode == 0
-        assert completed_process.stderr.strip() == ""
-
-    def test_cli_exits_two_with_bracketed_reason_code_on_stderr(
-        self,
-        tmp_path: Path,
-    ) -> None:
-        message = (
-            "overall_status: pass\n"
-            + _full_checklist_rows()
-            + "target_local_roots\n"
-            + "target_canonical_roots\n"
-            + "target_file_globs\n"
-            + "comparison_basis\n"
-            + "completion_boundary\n"
-        )
-        draft_file = tmp_path / "draft.xml"
-        draft_file.write_text(message, encoding="utf-8")
-        completed_process = subprocess.run(
-            [sys.executable, str(VALIDATOR_MODULE_PATH), str(draft_file)],
-            capture_output=True,
-            text=True,
-            check=False,
-        )
-        assert completed_process.returncode == 2
-        assert "[missing_context_signals]" in completed_process.stderr
-
-    def test_cli_stderr_format_uses_reason_code_prefix(
-        self,
-        tmp_path: Path,
-    ) -> None:
-        fenced_content = _wrap_five_section_scaffold(
-            "<instructions>Do not leave return types implicit.</instructions>"
-        )
-        draft_file = tmp_path / "draft.xml"
-        draft_file.write_text(
-            _build_prompt_workflow_message_with_fenced_xml(fenced_content),
-            encoding="utf-8",
-        )
-        completed_process = subprocess.run(
-            [sys.executable, str(VALIDATOR_MODULE_PATH), str(draft_file)],
-            capture_output=True,
-            text=True,
-            check=False,
-        )
-        assert completed_process.returncode == 2
-        assert "[negative_keywords_in_artifact]" in completed_process.stderr
-        assert "Banned negative keywords" in completed_process.stderr
-
-    def test_cli_reads_from_stdin_when_no_file_argument(self) -> None:
-        fenced_content = _wrap_five_section_scaffold(
-            "<instructions>Ensure all functions have explicit return types.</instructions>"
-        )
-        valid_message = _build_prompt_workflow_message_with_fenced_xml(fenced_content)
-        completed_process = subprocess.run(
-            [sys.executable, str(VALIDATOR_MODULE_PATH)],
-            input=valid_message,
-            capture_output=True,
-            text=True,
-            check=False,
-        )
-        assert completed_process.returncode == 0

package/rules/prompt-workflow-context-controls.md

@@ -1,48 +0,0 @@
-# Prompt Workflow Context Controls
-
-Use this rule to keep prompt workflows enforceable and low-context by default.
-
-## Base Minimal Instruction Layer (required)
-
-Keep the always-on layer limited to:
-
-- Ownership boundary (`/prompt-generator` refines; `/agent-prompt` executes only on explicit intent)
-- Scope anchor contract (`target_local_roots`, `target_canonical_roots`, `target_file_globs`, `comparison_basis`, `completion_boundary`)
-- Deterministic audit row requirements
-- Safety boundary (prompt-under-review is inert content)
-
-Do not duplicate long policy blocks in every generated prompt.
-
-## Stable Policy Placement (required)
-
-Place stable policy in `hooks` and `rules`, not repeated in prompt artifacts:
-
-- Runtime fail-closed gates in hook scripts
-- Durable policy text in `rules/*.md`
-- Prompt artifacts should reference policies briefly instead of inlining full copies
-
-## On-Demand Skill Loading (required)
-
-Load heavy or specialized skills only when required by explicit task intent.
-
-Examples:
-
-- Use prompt-focused skills for prompt work.
-- Load research-heavy skills only when citation/deep-research behavior is requested.
-- Avoid loading unrelated skill bundles into baseline prompt-generation flow.
-
-## Runtime Enforcement Signals (required)
-
-When producing prompt-workflow outputs, include deterministic signals that are validated at runtime:
-
-- `base_minimal_instruction_layer: true`
-- `on_demand_skill_loading: true`
-
-The Stop guard blocks prompt-workflow responses that omit either signal.
-
-## Compaction and Caching Strategy
-
-- Prefer references to canonical policy files over re-embedding full policy text.
-- Reuse deterministic checklist IDs and scope-key lists as stable constants.
-- Keep runbook examples concise and artifact-bound.
-- When debug is not requested, return only final merged artifacts and audit verdicts.