prizmkit 1.1.72 → 1.1.76

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

@@ -137,7 +137,10 @@ class ProgressTracker:
         self.event_format = ""
         self.active_subagent_count = 0
         self.subagent_status_counts = Counter()
+        self._subagent_spawn_count = 0
         self.codex_child_thread_ids = set()
+        self.cb_session_id = ""
+        self.cb_cwd = ""
         self.claude_session_id = ""
         self.claude_cwd = ""
         self.claude_task_states = {}
@@ -147,6 +150,7 @@ class ProgressTracker:
         self.last_child_activity_at = ""
         self._codex_child_session_paths = {}
         self._claude_child_session_files = []
+        self._cb_child_session_files = []
         self._last_child_scan_at = 0.0
         self._last_claude_fallback_scan_at = 0.0
         self._last_claude_fallback_scan_key = ""
@@ -158,10 +162,13 @@ class ProgressTracker:
     def process_event(self, event):
         """Process a single stream-json event and update state.
 
-        Supports two formats:
+        Supports three formats:
         1. Claude API raw stream: message_start, content_block_start, content_block_delta, etc.
         2. Claude Code verbose stream-json: assistant, user, tool_result, system, etc.
            (produced by claude/claude-internal --verbose --output-format stream-json)
+        3. CodeBuddy stream-json: message/function_call/function_call_result events
+           (produced by cbc --print -y --output-format stream-json — message type
+           with sessionId/cwd metadata, function_call for tool invocations).
         """
         event_type = event.get("type", "")
 
@@ -233,6 +240,93 @@ class ProgressTracker:
 
             return
 
+        # ── CodeBuddy stream-json format ──────────────────────────────
+        # cbc --print -y --output-format stream-json emits:
+        #   message (role=user/assistant), function_call, function_call_result,
+        #   file-history-snapshot, error.  The first user message carries
+        #   sessionId and cwd metadata; later function_call items carry the
+        #   same fields.
+        # NOTE: cbc emits tool invocations as EITHER tool_use blocks inside a
+        # message (role=assistant) event OR as standalone function_call events,
+        # but not both for the same invocation.  The two handlers are
+        # complementary rather than redundant, so no deduplication is needed.
+        # Detect via event_type=="message" with a "sessionId" field.
+        # Must run before the Claude Code assistant / tool_result branches
+        # because those also accept those types but use different internals.
+        if event_type == "message" and isinstance(event.get("sessionId"), str):
+            self.event_format = self.event_format or "stream-json"
+            role = event.get("role", "")
+            sid = event.get("sessionId", "")
+            cwd = event.get("cwd", "")
+            if sid and not self.cb_session_id:
+                self.cb_session_id = sid.strip()
+            if cwd and not self.cb_cwd:
+                self.cb_cwd = cwd.strip()
+            if role == "assistant":
+                self.message_count += 1
+                self.is_active = True
+                content = event.get("content", [])
+                if isinstance(content, list):
+                    for block in content:
+                        block_type = block.get("type", "")
+                        if block_type == "tool_use":
+                            tool_name = block.get("name", "unknown")
+                            self.current_tool = tool_name
+                            self.tool_call_counts[tool_name] += 1
+                            self.total_tool_calls += 1
+                            tool_input = block.get("input", {})
+                            if isinstance(tool_input, dict):
+                                self._extract_tool_summary_from_dict(tool_input)
+                            self._detect_phase(
+                                json.dumps(tool_input, ensure_ascii=False)[:500]
+                            )
+                            # Track CodeBuddy Agent/Task tool invocations as
+                            # potential sub-agent spawns (Gap 3 fix).
+                            self._record_cb_agent_tool_call(tool_name, tool_input)
+                        elif block_type == "text":
+                            text = block.get("text", "")
+                            if text.strip():
+                                self.last_text_snippet = text.strip()[:120]
+                            self._detect_phase(text)
+                            self._detect_terminal_error(
+                                text, require_error_context=True
+                            )
+            return
+
+        # CodeBuddy function_call / function_call_result in stream-json
+        if event_type in ("function_call", "function_call_result"):
+            self.event_format = self.event_format or "stream-json"
+            sid = event.get("sessionId", "")
+            cwd = event.get("cwd", "")
+            if sid and not self.cb_session_id:
+                self.cb_session_id = sid.strip()
+            if cwd and not self.cb_cwd:
+                self.cb_cwd = cwd.strip()
+            if event_type == "function_call":
+                tool_name = event.get("name", "unknown")
+                self.current_tool = tool_name
+                self.tool_call_counts[tool_name] += 1
+                self.total_tool_calls += 1
+                self.is_active = True
+                # Extract summary from arguments
+                raw_args = event.get("arguments", "")
+                if isinstance(raw_args, str) and raw_args.strip():
+                    self._extract_tool_summary(raw_args)
+                    self._detect_phase(raw_args[:500])
+                    # Track Agent/Task tool invocations as sub-agent spawns
+                    self._record_cb_agent_tool_call(tool_name, raw_args)
+            elif event_type == "function_call_result":
+                self.current_tool = None
+                # Check result text for phase / error hints
+                result_text = event.get("output") or ""
+                if result_text and isinstance(result_text, str):
+                    stripped = result_text.strip()
+                    if stripped:
+                        self.last_text_snippet = stripped[:120]
+                    self._detect_phase(stripped)
+                    self._detect_terminal_error(stripped, require_error_context=True)
+            return
+
         # ── Claude Code verbose format ──────────────────────────────
         if event_type == "assistant":
             self.event_format = self.event_format or "stream-json"
@@ -259,11 +353,52 @@ class ProgressTracker:
                     self._detect_phase(text)
                     self._detect_terminal_error(text, require_error_context=True)
 
-        elif event_type == "tool_result" or event_type == "user":
+        elif event_type == "tool_result" or (
+            event_type == "user"
+            and not isinstance(event.get("sessionId"), str)
+        ):
             # 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 (Claude Code)
+            if event_type == "tool_result":
+                content_candidates.append(str(event.get("content", "")))
+
+            # Format B: nested inside user event (Claude Code, NOT CodeBuddy)
+            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"
@@ -427,15 +562,60 @@ class ProgressTracker:
             self._current_tool_input_parts = []
 
         elif event_type == "error":
+            self.event_format = self.event_format or "stream-json"
             error_msg = event.get("error", {}).get("message", "Unknown error")
+            errors = event.get("errors") or []
+            if isinstance(errors, list) and errors:
+                error_msg = "; ".join(str(e) for e in errors[:3])
             self.errors.append(error_msg)
             self._detect_terminal_error(str(error_msg))
 
+        # ── CodeBuddy file-history-snapshot (ignore) ─────────────────
+        elif event_type == "file-history-snapshot":
+            return
+
         # 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_cb_agent_tool_call(self, tool_name, raw_args):
+        """Record a CodeBuddy Agent/Task* tool invocation for sub-agent tracking.
+
+        CodeBuddy spawns sub-agents via:
+        - Agent(prompt=..., run_in_background=True/False) — synchronous or bg fork
+        - TaskCreate(subject=..., description=...) — scheduled task
+
+        (TaskUpdate/TaskOutput exist but are lifecycle-only — they track
+        existing tasks rather than creating new sub-agents, so we don't
+        count them as spawns.)
+
+        We don't get child session ids from these tool calls in the stream,
+        but recording the count lets the heartbeat code in heartbeat.sh apply
+        an extended stale-kill window just as it does for Codex spawn_agent.
+        """
+        if tool_name not in ("Agent", "TaskCreate"):
+            return
+        # Both dict (from message events) and str/JSON (from function_call events)
+        # are supported.
+        if isinstance(raw_args, dict):
+            args = raw_args
+        elif isinstance(raw_args, str):
+            try:
+                args = json.loads(raw_args)
+            except (json.JSONDecodeError, TypeError):
+                return
+        else:
+            return
+        # For Agent tool, subagent_type or run_in_background hints at delegation
+        if tool_name == "Agent":
+            subagent_type = args.get("subagent_type", "")
+            prompt = args.get("prompt", "")
+            if subagent_type or prompt:
+                self._subagent_spawn_count += 1
+        elif tool_name == "TaskCreate":
+            self._subagent_spawn_count += 1
+
     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 "")
@@ -719,17 +899,85 @@ class ProgressTracker:
             return []
         return matches
 
+    def _cb_projects_dir(self):
+        """Return the CodeBuddy projects directory for transcript lookup.
+
+        CodeBuddy stores per-project session transcripts and sub-agent data
+        under ~/.codebuddy/projects/{projectDir}/{sessionId}/.
+        """
+        cb_config = os.environ.get("CODEBUDDY_CONFIG_DIR")
+        if cb_config:
+            return Path(cb_config).expanduser() / "projects"
+        cb_home = os.environ.get("CODEBUDDY_HOME")
+        if cb_home:
+            return Path(cb_home).expanduser() / "projects"
+        return Path.home() / ".codebuddy" / "projects"
+
+    def _cb_project_key(self):
+        """Encode cwd the same way CodeBuddy stores project subdirs.
+
+        CodeBuddy uses the same sanitisation as Claude Code (\, /, : → -)
+        but strips the leading '-' so "/Users/test/MyProject" becomes
+        "Users-test-MyProject".
+        """
+        cwd = self.cb_cwd
+        if not cwd:
+            return ""
+        return cwd.replace("\\", "-").replace("/", "-").replace(":", "").lstrip("-")
+
+    def _find_cb_child_session_files(self):
+        """Find CodeBuddy subagent transcript data for this parent session.
+
+        CodeBuddy sub-agents store tool-results/{callId}.txt files; conversation
+        transcripts are not (as of 2026) written as agent-*.jsonl files in the
+        subagents/ directory.  We track the tool-results .txt files as a proxy
+        for child activity so the heartbeat monitor can extend the stale-kill
+        window while sub-agents are running.
+        """
+        if not self.cb_session_id:
+            return []
+
+        projects_dir = self._cb_projects_dir()
+        if not projects_dir.exists():
+            return []
+
+        project_key = self._cb_project_key()
+        if not project_key:
+            return []
+
+        subagents_dir = (
+            projects_dir / project_key / self.cb_session_id / "subagents"
+        )
+        if not subagents_dir.exists():
+            return []
+
+        # Collect all files under each agent-{id} subdirectory. These are
+        # currently tool-results/{callId}.txt files, but the glob also picks
+        # up future agent-*.jsonl transcripts should CodeBuddy add them.
+        try:
+            return sorted(
+                p for p in subagents_dir.glob("**/*")
+                if p.is_file()
+            )
+        except OSError:
+            return []
+
     def refresh_child_session_activity(self, force=False):
         """Refresh child transcript file stats.
 
         The heartbeat monitor uses this activity signature to treat subagent
         transcript growth as real progress while the parent session is blocked
-        waiting for a child agent/tool result. Supports Codex child threads and
-        Claude Code in-process teammate transcripts.
+        waiting for a child agent/tool result. Supports Codex child threads,
+        Claude Code in-process teammate transcripts, and CodeBuddy sub-agent
+        file activity under subagents/.
         """
         previous_signature = self.child_activity_signature
 
-        if not self.codex_child_thread_ids and not self.claude_session_id:
+        if (
+            not self.codex_child_thread_ids
+            and not self.claude_session_id
+            and not self.cb_session_id
+        ):
             self.child_session_files = []
             self.child_total_bytes = 0
             self.child_activity_signature = ""
@@ -750,6 +998,7 @@ class ProgressTracker:
                     if found:
                         self._codex_child_session_paths[thread_id] = found
             self._claude_child_session_files = self._find_claude_child_session_files()
+            self._cb_child_session_files = self._find_cb_child_session_files()
             self._last_child_scan_at = now
 
         files = []
@@ -788,6 +1037,14 @@ class ProgressTracker:
         for path in self._claude_child_session_files:
             add_file("claude", path.stem, path)
 
+        for path in self._cb_child_session_files:
+            # Identifier for CodeBuddy sub-agent files: use the parent
+            # directory name (agent-{id}) so heartbeat can distinguish
+            # activity across different sub-agent instances.
+            parent_name = path.parent.name if hasattr(path, "parent") else ""
+            identifier = parent_name if parent_name.startswith("agent-") else path.name
+            add_file("codebuddy", identifier, path)
+
         self.child_session_files = files
         self.child_total_bytes = total_bytes
         self.child_activity_signature = "|".join(signature_parts)
@@ -823,11 +1080,14 @@ class ProgressTracker:
             "total_tool_calls": self.total_tool_calls,
             "active_subagent_count": self.active_subagent_count,
             "subagent_states": subagent_states,
+            "subagent_spawn_count": self._subagent_spawn_count,
             "child_thread_ids": sorted(self.codex_child_thread_ids),
             "child_session_files": self.child_session_files,
             "child_total_bytes": self.child_total_bytes,
             "child_activity_signature": self.child_activity_signature,
             "last_child_activity_at": self.last_child_activity_at,
+            "cb_session_id": self.cb_session_id,
+            "cb_cwd": self.cb_cwd,
             "last_text_snippet": self.last_text_snippet,
             "last_result_is_error": self.last_result_is_error,
             "api_error_status": self.api_error_status,

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
+```