prizmkit 1.1.70 → 1.1.74

package/bundled/dev-pipeline-windows/scripts/parse-stream-progress.py

@@ -17,6 +17,7 @@ The script runs until:
 import argparse
 import json
 import os
+import re
 import signal
 import sys
 import tempfile
@@ -59,6 +60,58 @@ PHASE_KEYWORDS = {
     },
 }
 
+CONTEXT_ERROR_PATTERNS = [
+    re.compile(pattern, re.IGNORECASE)
+    for pattern in (
+        r"context_too_large",
+        r"model_context_window_exceeded",
+        r"Your input exceeds the context window",
+        r"input exceeds the context window",
+        r"context window of this model",
+        r"context window exceeded",
+        r"invalid_request_error.*context window",
+        r"context window.*invalid_request_error",
+    )
+]
+
+ERROR_CONTEXT_PATTERNS = [
+    re.compile(pattern, re.IGNORECASE)
+    for pattern in (
+        r"\bapi error\b",
+        r"invalid_request_error",
+        r"\bstatus\s*[:=]?\s*(400|413)\b",
+        r"\bapi_error_status\b",
+        r"\bapi_error_code\b",
+        r"\blast_result_is_error\b\s*[\"':=]*\s*true\b",
+        r"\bis_error\b\s*[\"':=]*\s*true\b",
+    )
+]
+
+
+def _has_error_context(text):
+    """Return true when free text looks like a runtime/provider error."""
+    if not text:
+        return False
+    return any(pattern.search(text) for pattern in ERROR_CONTEXT_PATTERNS)
+
+
+def detect_api_error_code(text, require_error_context=False):
+    """Return a normalized fatal/runtime error code from terminal text.
+
+    Structured terminal result/error events and raw stderr can be matched
+    directly. Ordinary assistant prose is noisier: it may mention the phrase
+    "input exceeds the context window" while explaining a test or recovery
+    rule, so callers can require additional error-like context there.
+    """
+    if not text:
+        return ""
+    if require_error_context and not _has_error_context(text):
+        return ""
+    for pattern in CONTEXT_ERROR_PATTERNS:
+        if pattern.search(text):
+            return "context_too_large"
+    return ""
+
 
 class ProgressTracker:
     """Tracks progress state from stream-json events."""
@@ -73,6 +126,12 @@ class ProgressTracker:
         self.tool_call_counts = Counter()
         self.total_tool_calls = 0
         self.last_text_snippet = ""
+        self.last_result_is_error = False
+        self.api_error_status = None
+        self.api_error_code = ""
+        self.terminal_result_text = ""
+        self.terminal_success_at = ""
+        self.fatal_error_code = ""
         self.is_active = True
         self.errors = []
         self.event_format = ""
@@ -164,11 +223,13 @@ class ProgressTracker:
             elif event_type == "turn.failed":
                 error = event.get("error") or event.get("message") or "Codex turn failed"
                 self.errors.append(str(error))
+                self._detect_terminal_error(str(error))
                 self.current_tool = None
 
             elif event_type == "error":
                 error = event.get("error") or event.get("message") or "Unknown error"
                 self.errors.append(str(error))
+                self._detect_terminal_error(str(error))
 
             return
 
@@ -196,12 +257,51 @@ class ProgressTracker:
                     if text.strip():
                         self.last_text_snippet = text.strip()[:120]
                     self._detect_phase(text)
+                    self._detect_terminal_error(text, require_error_context=True)
 
         elif event_type == "tool_result" or event_type == "user":
             # tool_result contains output from tool execution
             self.event_format = self.event_format or "stream-json"
             self.is_active = True
 
+            # Check for error patterns in tool_result content (supports both formats):
+            # A) Top-level tool_result events: event["content"] is the result text
+            # B) Nested user events: event["message"]["content"][] has type=="tool_result" items
+            content_candidates = []
+
+            # Format A: top-level tool_result
+            if event_type == "tool_result":
+                content_candidates.append(str(event.get("content", "")))
+
+            # Format B: nested inside user event
+            if event_type == "user":
+                message = event.get("message", {})
+                content_list = message.get("content", [])
+                if isinstance(content_list, list):
+                    for item in content_list:
+                        if isinstance(item, dict) and item.get("type") == "tool_result":
+                            content_candidates.append(str(item.get("content", "")))
+
+            for result_text in content_candidates:
+                if "shorter than the provided offset" in result_text:
+                    self.errors.append({
+                        "type": "read_offset_overflow",
+                        "tool": self.current_tool,
+                        "at": datetime.now(timezone.utc).isoformat(),
+                    })
+                    break  # one error per event is enough
+                elif "Wasted call" in result_text:
+                    self.errors.append({
+                        "type": "wasted_call",
+                        "tool": self.current_tool,
+                        "at": datetime.now(timezone.utc).isoformat(),
+                    })
+                    break
+
+            # Keep only last 20 errors to prevent unbounded growth in progress.json
+            if len(self.errors) > 20:
+                self.errors = self.errors[-20:]
+
         elif event_type == "system":
             # System events (hooks, init, task notifications, etc.) — track but don't count as messages.
             self.event_format = self.event_format or "stream-json"
@@ -274,6 +374,28 @@ class ProgressTracker:
                         state.setdefault("subagent_type", "")
                     self._update_claude_subagent_status_counts()
 
+        elif event_type == "result":
+            self.event_format = self.event_format or "stream-json"
+            self.is_active = False
+            result_text = event.get("result") or event.get("message") or ""
+            error_obj = event.get("error")
+            if isinstance(error_obj, dict):
+                error_text = " ".join(
+                    str(error_obj.get(key) or "")
+                    for key in ("type", "code", "message")
+                    if error_obj.get(key)
+                )
+                result_text = " ".join(part for part in (str(result_text), error_text) if part)
+            api_error_code = event.get("api_error_code") or event.get("error_code") or ""
+            if isinstance(error_obj, dict) and not api_error_code:
+                api_error_code = error_obj.get("code") or error_obj.get("type") or ""
+            self._record_terminal_result(
+                text=str(result_text or ""),
+                is_error=bool(event.get("is_error")),
+                api_error_status=event.get("api_error_status"),
+                api_error_code=str(api_error_code or ""),
+            )
+
         # ── Claude API raw stream format ────────────────────────────
         elif event_type == "message_start":
             self.event_format = self.event_format or "stream-json"
@@ -316,6 +438,7 @@ class ProgressTracker:
                     self.last_text_snippet = stripped[:120]
                 # Try to detect phase from text
                 self._detect_phase(text)
+                self._detect_terminal_error(text, require_error_context=True)
 
             elif delta_type == "input_json_delta":
                 partial = delta.get("partial_json", "")
@@ -331,21 +454,73 @@ class ProgressTracker:
                 self._extract_tool_summary(full_input)
                 self._detect_phase(full_input)
             else:
-                # Text block finished - detect phase from accumulated text
+                # Text block finished - detect phase and terminal errors from accumulated text
                 if self._text_buffer:
                     self._detect_phase(self._text_buffer)
+                    self._detect_terminal_error(
+                        self._text_buffer,
+                        require_error_context=True,
+                    )
             self._in_tool_use = False
             self._current_tool_input_parts = []
 
         elif event_type == "error":
             error_msg = event.get("error", {}).get("message", "Unknown error")
             self.errors.append(error_msg)
+            self._detect_terminal_error(str(error_msg))
 
         # Check for subagent indicator
         if event.get("parent_tool_use_id"):
             # This is a sub-agent event; tool name is still tracked normally
             pass
 
+    def _record_terminal_result(self, text="", is_error=False, api_error_status=None, api_error_code=""):
+        """Record a Claude Code terminal result event."""
+        terminal_text = str(text or "")
+        self.last_result_is_error = bool(is_error)
+        if api_error_status not in (None, ""):
+            try:
+                self.api_error_status = int(api_error_status)
+            except (TypeError, ValueError):
+                self.api_error_status = api_error_status
+        error_like_result = (
+            self.last_result_is_error
+            or api_error_status not in (None, "")
+            or bool(api_error_code)
+            or _has_error_context(terminal_text)
+        )
+        normalized_code = detect_api_error_code(
+            " ".join([str(api_error_code or ""), terminal_text]),
+            require_error_context=not error_like_result,
+        )
+        if normalized_code:
+            self.api_error_code = normalized_code
+            self.fatal_error_code = normalized_code
+        elif api_error_code:
+            self.api_error_code = str(api_error_code)
+        self.terminal_result_text = terminal_text[:1000]
+        if terminal_text.strip():
+            self.last_text_snippet = terminal_text.strip()[:120]
+        if not self.last_result_is_error and not self.fatal_error_code:
+            self.terminal_success_at = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
+        elif self.fatal_error_code:
+            self.errors.append(self.fatal_error_code)
+
+    def _detect_terminal_error(self, text, require_error_context=False):
+        """Detect fatal context-window errors from unstructured text."""
+        code = detect_api_error_code(
+            str(text or ""),
+            require_error_context=require_error_context,
+        )
+        if not code:
+            return
+        self.last_result_is_error = True
+        self.api_error_code = code
+        self.fatal_error_code = code
+        self.terminal_result_text = str(text or "")[:1000]
+        if text:
+            self.last_text_snippet = str(text).strip()[:120]
+
     def _detect_phase(self, text):
         """Detect pipeline phase from text content.
 
@@ -692,6 +867,12 @@ class ProgressTracker:
             "child_activity_signature": self.child_activity_signature,
             "last_child_activity_at": self.last_child_activity_at,
             "last_text_snippet": self.last_text_snippet,
+            "last_result_is_error": self.last_result_is_error,
+            "api_error_status": self.api_error_status,
+            "api_error_code": self.api_error_code,
+            "terminal_result_text": self.terminal_result_text,
+            "terminal_success_at": self.terminal_success_at,
+            "fatal_error_code": self.fatal_error_code,
             "is_active": self.is_active,
             "errors": self.errors[-10:],  # Keep last 10 errors
         }
@@ -728,6 +909,12 @@ def tail_and_parse(session_log, progress_file, poll_interval=0.5):
             state["current_phase"],
             state["total_tool_calls"],
             state.get("child_activity_signature", ""),
+            state.get("last_result_is_error"),
+            state.get("api_error_status"),
+            state.get("api_error_code", ""),
+            state.get("fatal_error_code", ""),
+            state.get("terminal_result_text", ""),
+            tuple(state.get("errors", [])),
         )
 
     # Wait for log file to appear
@@ -752,11 +939,19 @@ def tail_and_parse(session_log, progress_file, poll_interval=0.5):
                     event = json.loads(line)
                     tracker.process_event(event)
                 except json.JSONDecodeError:
-                    # Not a JSON line (could be stderr mixed in)
-                    # Use it as a text snippet if meaningful
+                    # Not a JSON line (could be stderr mixed in). Use it as a
+                    # text snippet and only treat it as terminal when it has a
+                    # strong API/runtime error marker; ordinary assistant prose
+                    # can discuss context limits without being fatal.
                     stripped = line.strip()
                     if stripped and len(stripped) > 5:
                         tracker.last_text_snippet = stripped[:120]
+                        tracker._detect_terminal_error(stripped, require_error_context=True)
+                        current_state = tracker.to_dict()
+                        current_state_key = state_key(current_state)
+                        if current_state_key != last_write_state:
+                            atomic_write_json(current_state, progress_file)
+                            last_write_state = current_state_key
                     continue
 
                 # Write progress if state changed

package/bundled/dev-pipeline-windows/scripts/update-feature-status.py

@@ -49,6 +49,7 @@ SESSION_STATUS_VALUES = [
     "commit_missing",
     "docs_missing",
     "merge_conflict",
+    "finalization_needed",
 ]
 
 TERMINAL_STATUSES = {"completed", "failed", "skipped", "auto_skipped", "split"}
@@ -644,7 +645,25 @@ def action_update(args, feature_list_path, state_dir):
         fs["degraded_reason"] = session_status
         fs["resume_from_phase"] = None
         fs["sessions"] = []
-        fs["last_session_id"] = None
+        if session_id:
+            fs["last_session_id"] = session_id
+            fs["last_failed_session_id"] = session_id
+
+        err = update_feature_in_list(feature_list_path, feature_id, new_status)
+        if err:
+            error_out("Failed to update .prizmkit/plans/feature-list.json: {}".format(err))
+            return
+    elif session_status == "finalization_needed":
+        # Runtime preserved dirty post-completion changes but could not safely
+        # clean them for automatic merge. Preserve the dev branch and stop for
+        # manual finalization instead of spending code retry budget.
+        new_status = "failed"
+        fs["degraded_reason"] = session_status
+        fs["resume_from_phase"] = None
+        fs["finalization_needed_count"] = fs.get("finalization_needed_count", 0) + 1
+        if session_id:
+            fs["last_session_id"] = session_id
+            fs["last_failed_session_id"] = session_id
 
         err = update_feature_in_list(feature_list_path, feature_id, new_status)
         if err:
@@ -657,6 +676,8 @@ def action_update(args, feature_list_path, state_dir):
         new_status = "pending"
         fs["infra_error_count"] = fs.get("infra_error_count", 0) + 1
         fs["last_infra_error_session_id"] = session_id
+        if session_id:
+            fs["last_session_id"] = session_id
         fs["resume_from_phase"] = None
 
         err = update_feature_in_list(feature_list_path, feature_id, new_status)
@@ -673,6 +694,9 @@ def action_update(args, feature_list_path, state_dir):
             new_status = "pending"
 
         fs["resume_from_phase"] = None
+        if session_id:
+            fs["last_session_id"] = session_id
+            fs["last_failed_session_id"] = session_id
         # Keep sessions list and last_session_id for debugging
 
         err = update_feature_in_list(feature_list_path, feature_id, new_status)
@@ -712,9 +736,9 @@ def action_update(args, feature_list_path, state_dir):
     }
     if auto_skipped_features:
         summary["auto_skipped"] = [info["feature_id"] for info in auto_skipped_features]
-    if session_status in ("commit_missing", "docs_missing", "merge_conflict"):
+    if session_status in ("commit_missing", "docs_missing", "merge_conflict", "finalization_needed"):
         summary["degraded_reason"] = session_status
-        summary["restart_policy"] = "finalization_retry"
+        summary["restart_policy"] = "manual_finalization" if session_status == "finalization_needed" else "finalization_retry"
     elif session_status == "infra_error":
         summary["restart_policy"] = "infra_retry"
         summary["infra_error_count"] = fs.get("infra_error_count", 0)

package/bundled/dev-pipeline-windows/templates/agent-prompts/dev-implement.md

@@ -1,5 +1,23 @@
 "Read {{DEV_SUBAGENT_PATH}}. Implement feature {{FEATURE_ID}} (slug: {{FEATURE_SLUG}}).
 
+## Task Summary Card
+
+**Objective**: Implement {{FEATURE_TITLE}}.
+
+**Primary files** (see context-snapshot.md Section 4 for full manifest):
+- Review plan.md Tasks section for the complete task-to-file mapping.
+- Each task's `— file:` suffix identifies the target file.
+
+**Test command**: `{{TEST_CMD}}`
+
+**Known baseline failures**: `{{BASELINE_FAILURES}}`
+
+**DO NOT**:
+- Re-read source files already listed in context-snapshot.md Section 4 File Manifest
+- Create new files unless a plan.md task explicitly requires it
+- Run git commands
+- Use mock success data or fake rows in UI/tests
+
 ## Required Inputs
 
 1. Read `.prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md` first.
@@ -35,6 +53,9 @@ Before returning, append `## Implementation Log` to `context-snapshot.md` with:
 - Carry forward the Dev-isolated subset: skip scaffold/generated files listed in `context-snapshot.md`; verify dependency versions before install/build commands that resolve dependencies; after build/compile commands, ensure outputs are ignored and never commit generated artifacts.
 - If tests fail, follow this Test Failure Recovery subset: classify failures as baseline, new regression, brittle test, or environment/tooling; fix new regressions and brittle tests while progress is being made; document baseline failures; write `failure-log.md` for blockers.
 - Do not run git commands; staging and commit are handled by the orchestrator.
+- **Edit safety**: If an Edit fails with 'String to replace not found', use Select-String or a targeted Read to find the target text before retrying. Never guess file offsets — verify them first.
+- **Read safety**: If 3 consecutive Reads to the same file return 'shorter than offset' or 'Wasted call', STOP and report BLOCKED.
+- **Test early**: Run `{{TEST_CMD}}` after every 3 successful Edit operations. Capture output with `Tee-Object (Join-Path $env:TEMP "test-out.txt")` and use Select-String for failures.
 
 Do not return success unless:
 1. implementation tasks are complete;

package/bundled/dev-pipeline-windows/templates/agent-prompts/reviewer-review.md

@@ -2,6 +2,6 @@
 1. Read `.prizmkit/specs/{{FEATURE_SLUG}}/spec.md` (if it exists) for goals and Verification Gates; if spec.md does not exist, read `.prizmkit/specs/{{FEATURE_SLUG}}/context-snapshot.md` Section 1 Task Contract instead
 2. Read `.prizmkit/specs/{{FEATURE_SLUG}}/plan.md` for architecture decisions and completed tasks
 3. Run /prizmkit-code-review with artifact_dir=.prizmkit/specs/{{FEATURE_SLUG}}/. The skill will run its internal review-fix loop (Reviewer → filter → Dev fix, max 3 rounds) and write review-report.md.
-4. Run the full test suite using `{{TEST_CMD}}`. When running tests: `({{TEST_CMD}}) 2>&1 | Tee-Object (Join-Path $env:TEMP "review-test-out.txt") | Select-Object -Last 20`, then Select-String (Join-Path $env:TEMP "review-test-out.txt") for details — do NOT re-run the suite multiple times.
+4. Run the full test suite using `{{TEST_CMD}}`. When running tests: `& { {{TEST_CMD}} } 2>&1 | Tee-Object (Join-Path $env:TEMP "review-test-out.txt") | Select-Object -Last 20`, then Select-String `(Join-Path $env:TEMP "review-test-out.txt")` for details — do NOT re-run the suite multiple times.
 5. review-report.md will be written to .prizmkit/specs/{{FEATURE_SLUG}}/ by prizmkit-code-review.
 Report: verdict (PASS/NEEDS_FIXES), number of rounds, findings fixed/rejected."

package/bundled/dev-pipeline-windows/templates/bootstrap-prompt.md

@@ -17,6 +17,12 @@
 
 {{ACCEPTANCE_CRITERIA}}
 
+## Your Mission
+
+You are the **session orchestrator**. Implement Feature {{FEATURE_ID}}: "{{FEATURE_TITLE}}".
+
+**CRITICAL**: You MUST NOT exit until ALL work is complete and committed.
+
 ## Instructions
 
 You are running in **headless non-interactive mode** — no human is available for input.
@@ -44,3 +50,24 @@ Infer what needs to be done from the feature context above and follow the standa
 - Do NOT push to remote — the user will push manually.
 - Write all artifacts to `.prizmkit/specs/{{FEATURE_SLUG}}/`.
 - If a step fails after 3 attempts, write a status report and stop.
+
+
+## Failure Capture Protocol
+
+If you encounter an unrecoverable error, context overflow, or are about to exit without completing all steps:
+
+1. Write `.prizmkit/specs/{{FEATURE_SLUG}}/failure-log.md` BEFORE exiting:
+   ```
+   FAILURE_TYPE: timeout | test_failure | review_rejected | context_overflow | unknown
+   PHASE: <failed phase>
+   ROOT_CAUSE: <1-2 sentence explanation>
+   ATTEMPTED: <approaches already tried>
+   SUGGESTION: <what the next session should try differently>
+   ```
+
+2. This file is intentionally lightweight — write it BEFORE context runs out.
+
+After a successful session (all steps complete + commit done), delete it:
+```powershell
+Remove-Item -Force -ErrorAction SilentlyContinue .prizmkit/specs/{{FEATURE_SLUG}}/failure-log.md
+```