prizmkit 1.1.140 → 1.1.142

package/bundled/dev-pipeline/scripts/generate-bugfix-prompt.py

@@ -40,15 +40,14 @@ def route_bugfix_repair(stage_result, repair_scope=None, repair_round=0):
     return route_pipeline_repair(stage_result, repair_scope, repair_round)
 
 
-def bugfix_checkpoint_metadata(artifact_dir, evidence=None, stage="plan",
-                               status="PLAN_READY", stage_result=None,
+def bugfix_checkpoint_metadata(artifact_dir, stage="plan",
+                               status="pending", stage_result=None,
                                repair_scope=None, repair_round=0,
                                next_stage="prizmkit-implement",
                                resume_from="prizmkit-implement"):
-    """Build Bugfix L4 handoff metadata while preserving Bugfix L1 state."""
+    """Build Bugfix pipeline handoff metadata while preserving skill state."""
     return pipeline_checkpoint_metadata(
         artifact_dir,
-        evidence=evidence,
         stage=stage,
         status=status,
         stage_result=stage_result,
@@ -71,9 +70,8 @@ BUGFIX_SECTION_TO_SKILL = {
     "bugfix-phase-test": checkpoint_step(
         "prizmkit-test", "Post-Review Test Gate",
         [
-            ".prizmkit/test/evidence/*/manifest.json",
-            ".prizmkit/test/evidence/*/verdict.json",
-            ".prizmkit/test/evidence/*/validation.json",
+            ".prizmkit/bugfix/{slug}/test-report.md",
+            ".prizmkit/bugfix/{slug}/test-result.json",
         ],
     ),
     "phase-browser": checkpoint_step("browser-verification", "Browser Verification", []),
@@ -538,7 +536,7 @@ def main():
     checkpoint["bugfix_state"] = bugfix_checkpoint_metadata(
         replacements["{{ARTIFACT_DIR}}"],
         stage="plan",
-        status="PLAN_READY",
+        status="pending",
         stage_result=None,
         repair_scope=None,
         repair_round=0,

package/bundled/dev-pipeline/scripts/generate-recovery-prompt.py

@@ -67,7 +67,7 @@ affected files, and root cause hypothesis.""",
         "Triage",
         """\
 Locate affected code and identify root cause:
-1. Read `.prizmkit/prizm-docs/root.prizm` then relevant L1/L2 docs for affected modules
+1. Read `.prizmkit/prizm-docs/root.prizm` then the relevant module indexes and detail docs for affected modules
 2. Read files mentioned in the bug description or error/stack trace
 3. Check `.prizmkit/prizm-docs/` TRAPS for known patterns
 4. Classify: root cause (confirmed/suspected), blast radius, fix complexity

package/bundled/dev-pipeline/scripts/generate-refactor-prompt.py

@@ -40,15 +40,14 @@ def route_refactor_repair(stage_result, repair_scope=None, repair_round=0):
     return route_pipeline_repair(stage_result, repair_scope, repair_round)
 
 
-def refactor_checkpoint_metadata(artifact_dir, evidence=None, stage="plan",
-                                 status="PLAN_READY", stage_result=None,
+def refactor_checkpoint_metadata(artifact_dir, stage="plan",
+                                 status="pending", stage_result=None,
                                  repair_scope=None, repair_round=0,
                                  next_stage="prizmkit-implement",
                                  resume_from="prizmkit-implement"):
-    """Build Refactor L4 handoff metadata while preserving Refactor L1 state."""
+    """Build Refactor pipeline handoff metadata while preserving skill state."""
     return pipeline_checkpoint_metadata(
         artifact_dir,
-        evidence=evidence,
         stage=stage,
         status=status,
         stage_result=stage_result,
@@ -76,9 +75,8 @@ REFACTOR_SECTION_TO_SKILL = {
     "refactor-phase-test": checkpoint_step(
         "prizmkit-test", "Post-Review Test Gate",
         [
-            ".prizmkit/test/evidence/*/manifest.json",
-            ".prizmkit/test/evidence/*/verdict.json",
-            ".prizmkit/test/evidence/*/validation.json",
+            ".prizmkit/refactor/{slug}/test-report.md",
+            ".prizmkit/refactor/{slug}/test-result.json",
         ],
     ),
     "phase-browser": checkpoint_step("browser-verification", "Browser Verification", []),
@@ -556,7 +554,7 @@ def main():
     checkpoint["refactor_state"] = refactor_checkpoint_metadata(
         replacements["{{ARTIFACT_DIR}}"],
         stage="plan",
-        status="PLAN_READY",
+        status="pending",
         stage_result=None,
         repair_scope=None,
         repair_round=0,

package/bundled/dev-pipeline/scripts/prompt_framework.py

@@ -8,20 +8,24 @@ so feature, bugfix, and refactor prompts do not drift.
 import glob
 import os
 import re
+import sys
+
+_PIPELINE_ROOT = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
+if _PIPELINE_ROOT not in sys.path:
+    sys.path.insert(0, _PIPELINE_ROOT)
+
+from prizmkit_runtime.checkpoint_state import (  # noqa: E402
+    VALID_CHECKPOINT_STATUSES,
+    expected_status_for_stage_result,
+)
+from prizmkit_runtime.test_result import valid_test_artifact_pair  # noqa: E402
 
 
 MAX_OUTER_REPAIR_ROUNDS = 3
-REPAIR_SCOPES_REQUIRING_REVIEW = {
-    "production",
-    "runtime",
-    "schema",
-    "dependency",
-    "public-interface",
-}
 
 
 def route_pipeline_repair(stage_result, repair_scope=None, repair_round=0):
-    """Map review/test outcomes to bounded, safe L4 continuation routes."""
+    """Map review/test outcomes to bounded, safe pipeline continuation routes."""
     try:
         current_round = int(repair_round)
     except (TypeError, ValueError):
@@ -46,19 +50,28 @@ def route_pipeline_repair(stage_result, repair_scope=None, repair_round=0):
         result.update(
             next_stage="prizmkit-retrospective",
             resume_from="prizmkit-retrospective",
-            reason="authoritative test evidence passed",
+            reason="project-native tests passed",
         )
         return result
     if stage_result == "TEST_BLOCKED":
         result.update(
             blocked=True,
             resumable=False,
-            reason="bounded automatic recovery cannot safely edit production",
+            reason="testing could not establish a safe verdict",
             blocked_reason="test_blocked",
             terminal_status="WORKFLOW_BLOCKED",
         )
         return result
-    if stage_result not in {"REVIEW_NEEDS_FIXES", "TEST_FAIL"}:
+    if stage_result == "TEST_NEEDS_FIXES":
+        result.update(
+            blocked=True,
+            resumable=False,
+            reason="prizmkit-test stopped with a known remaining correction",
+            blocked_reason="test_needs_fixes",
+            terminal_status="WORKFLOW_BLOCKED",
+        )
+        return result
+    if stage_result != "REVIEW_NEEDS_FIXES":
         result.update(
             blocked=True,
             reason="unknown stage result is unsafe",
@@ -76,55 +89,29 @@ def route_pipeline_repair(stage_result, repair_scope=None, repair_round=0):
         return result
 
     result["repair_round"] = current_round + 1
-    if stage_result == "REVIEW_NEEDS_FIXES":
-        result.update(
-            next_stage="prizmkit-implement",
-            resume_from="prizmkit-implement",
-            route=["prizmkit-implement", "prizmkit-code-review", "prizmkit-test"],
-            reason="review findings require implementation and fresh review",
-        )
-        return result
-    if repair_scope == "test-infrastructure":
-        result.update(
-            next_stage="prizmkit-implement",
-            resume_from="prizmkit-implement",
-            route=["prizmkit-implement", "prizmkit-test"],
-            reason="test infrastructure failure must not edit production speculatively",
-        )
-        return result
-    if repair_scope in REPAIR_SCOPES_REQUIRING_REVIEW:
-        result.update(
-            next_stage="prizmkit-implement",
-            resume_from="prizmkit-implement",
-            route=["prizmkit-implement", "prizmkit-code-review", "prizmkit-test"],
-            reason="production-facing failure requires review before retesting",
-        )
-        return result
     result.update(
-        blocked=True,
-        reason="unknown repair scope has no safe automatic route",
-        blocked_reason="unsafe_repair_scope",
-        terminal_status="WORKFLOW_BLOCKED",
+        next_stage="prizmkit-implement",
+        resume_from="prizmkit-implement",
+        route=["prizmkit-implement", "prizmkit-code-review", "prizmkit-test"],
+        reason="review findings require implementation and fresh review",
     )
     return result
 
 
-def pipeline_checkpoint_metadata(artifact_dir, evidence=None, stage="plan",
-                                 status="PLAN_READY", stage_result=None,
+def pipeline_checkpoint_metadata(artifact_dir, stage="plan",
+                                 status="pending", stage_result=None,
                                  repair_scope=None, repair_round=0,
                                  next_stage="prizmkit-implement",
                                  resume_from="prizmkit-implement",
                                  blocked_reason=None, terminal_status=None):
-    """Build family-neutral semantic L4 metadata without replacing L1 state."""
-    evidence = evidence or {}
-    if evidence:
-        evidence_paths = {
-            "manifest": evidence.get("manifest_path"),
-            "verdict": evidence.get("verdict_path"),
-            "validation": evidence.get("validation_path"),
-        }
-    else:
-        evidence_paths = {"manifest": None, "verdict": None, "validation": None}
+    """Build family-neutral pipeline metadata without replacing skill state."""
+    if status not in VALID_CHECKPOINT_STATUSES:
+        raise ValueError("Invalid checkpoint lifecycle status: {}".format(status))
+    expected_status = expected_status_for_stage_result(stage_result)
+    if expected_status is None and stage_result is not None:
+        raise ValueError("Invalid checkpoint stage result: {}".format(stage_result))
+    if expected_status is not None and status != expected_status:
+        raise ValueError("{} requires status={}".format(stage_result, expected_status))
     current_stage = None if terminal_status else stage
     semantic_next_stage = None if terminal_status else next_stage
     return {
@@ -137,7 +124,6 @@ def pipeline_checkpoint_metadata(artifact_dir, evidence=None, stage="plan",
         "stage_result": stage_result,
         "repair_scope": repair_scope,
         "repair_round": repair_round,
-        "authoritative_evidence_paths": evidence_paths,
         "next_stage": semantic_next_stage,
         "resume_from": resume_from,
         "blocked_reason": blocked_reason,
@@ -248,6 +234,29 @@ def completed_step_artifacts_valid(project_root, artifacts):
     return all(artifact_exists(project_root, artifact) for artifact in artifacts)
 
 
+def completed_test_artifacts_valid(project_root, artifacts):
+    """Validate a sibling test-report.md and TEST_PASS test-result.json pair."""
+    reports = [item for item in artifacts if item.endswith("/test-report.md")]
+    results = [item for item in artifacts if item.endswith("/test-result.json")]
+    if len(reports) != 1 or len(results) != 1:
+        return False
+    report_path = os.path.realpath(os.path.join(project_root, reports[0]))
+    result_path = os.path.realpath(os.path.join(project_root, results[0]))
+    root_path = os.path.realpath(project_root)
+    try:
+        if os.path.commonpath([root_path, report_path]) != root_path:
+            return False
+        if os.path.commonpath([root_path, result_path]) != root_path:
+            return False
+    except ValueError:
+        return False
+    return valid_test_artifact_pair(
+        report_path,
+        result_path,
+        required_result="TEST_PASS",
+    )
+
+
 def checkpoint_skip():
     """Return an explicit no-checkpoint-step contract for a prompt section."""
     return {"kind": "skip"}
@@ -360,7 +369,13 @@ def generate_checkpoint_from_sections(
 
 def merge_checkpoint_state(existing, fresh, project_root, validator=None, logger=None):
     """Merge existing checkpoint state into a freshly generated definition."""
-    validator = validator or (lambda _skill, artifacts: completed_step_artifacts_valid(project_root, artifacts))
+    validator = validator or (
+        lambda skill, artifacts: (
+            completed_test_artifacts_valid(project_root, artifacts)
+            if skill == "prizmkit-test"
+            else completed_step_artifacts_valid(project_root, artifacts)
+        )
+    )
     existing_by_id = {}
     existing_by_skill = {}
     for step in existing.get("steps", []):
@@ -402,9 +417,6 @@ def merge_checkpoint_state(existing, fresh, project_root, validator=None, logger
                 step["status"] = "completed"
                 if stage_result:
                     step["stage_result"] = stage_result
-                evidence = previous.get("authoritative_evidence_paths")
-                if isinstance(evidence, dict):
-                    step["authoritative_evidence_paths"] = evidence
                 continue
             if logger is not None:
                 logger.warning(

package/bundled/dev-pipeline/scripts/update-bug-status.py

@@ -47,6 +47,7 @@ SESSION_STATUS_VALUES = [
     "context_overflow",
     "stalled_context_continuation",
     "workflow_blocked",
+    "workflow_skipped",
     "commit_missing",
     "docs_missing",
     "merge_conflict",
@@ -89,7 +90,7 @@ def parse_args():
     parser.add_argument("--continuation-summary-path", default=None, help="Optional continuation summary artifact path.")
     parser.add_argument("--no-progress-count", type=int, default=None, help="Consecutive context-overflow sessions without objective progress.")
     parser.add_argument("--progress-fingerprint", default=None, help="JSON progress fingerprint captured after the session.")
-    parser.add_argument("--checkpoint-state", default=None, help="JSON semantic L4 checkpoint snapshot captured after the session.")
+    parser.add_argument("--checkpoint-state", default=None, help="JSON semantic pipeline checkpoint snapshot captured after the session.")
     return parser.parse_args()
 
 
@@ -261,7 +262,6 @@ SEMANTIC_CHECKPOINT_FIELDS = (
     "stage_result",
     "repair_scope",
     "repair_round",
-    "authoritative_evidence_paths",
     "blocked_reason",
     "terminal_status",
 )
@@ -594,6 +594,20 @@ def action_update(args, bug_list_path, state_dir):
             if err:
                 error_out("Failed to update .prizmkit/plans/bug-fix-list.json: {}".format(err))
                 return
+    elif session_status == "workflow_skipped":
+        # A completed AI CLI session without authoritative workflow evidence is
+        # skipped, not crashed. Preserve diagnostics and spend no retry budget.
+        new_status = "skipped"
+        bs["continuation_pending"] = False
+        bs.pop("continuation_reason", None)
+        bs.pop("terminal_status", None)
+        bs.pop("blocked_reason", None)
+        bs["needs_attention"] = True
+        bs["skipped_reason"] = "workflow_completion_evidence_invalid"
+        err = update_bug_in_list(bug_list_path, bug_id, new_status)
+        if err:
+            error_out("Failed to update .prizmkit/plans/bug-fix-list.json: {}".format(err))
+            return
     elif session_status == "workflow_blocked":
         # Semantic terminal state: preserve partial work without consuming any retry budget.
         new_status = "needs_info"
@@ -678,6 +692,11 @@ def action_update(args, bug_list_path, state_dir):
         summary["restart_policy"] = "context_overflow_continuation"
         summary.update(continuation_metadata_summary(bs))
         summary["artifacts_preserved"] = True
+    elif session_status == "workflow_skipped":
+        summary["restart_policy"] = "terminal_skipped"
+        summary.update(continuation_metadata_summary(bs))
+        summary["checkpoint_state"] = semantic_checkpoint
+        summary["artifacts_preserved"] = True
     elif session_status == "workflow_blocked":
         summary["restart_policy"] = "terminal_blocked"
         summary.update(continuation_metadata_summary(bs))