Merged
Conversation
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 3 potential issues.
Autofix Details
Bugbot Autofix prepared fixes for all 3 issues found in the latest run.
- ✅ Fixed: Substring "doc" falsely matches "docker" as docs archetype
- Replaced bare substring hint matching with compiled regex using word boundaries (
\bdocs?\b), so "doc"/"docs" match as standalone words but not inside "docker" or "docstring".
- Replaced bare substring hint matching with compiled regex using word boundaries (
- ✅ Fixed: Substring "port" falsely matches common words like "import"
- Replaced bare substring hint matching with compiled regex using word boundaries (
\bport\b), so "port" matches as a standalone word but not inside "import", "export", or "support".
- Replaced bare substring hint matching with compiled regex using word boundaries (
- ✅ Fixed: Verification run count tallies types not actual runs
- Changed
verification_run_countto iterate over the fullexecution_evidencelist instead of the deduplicatedexecution_typesset, so it counts actual verification runs rather than distinct type categories.
- Changed
Or push these changes by commenting:
@cursor push ffd74e90d1
Preview (ffd74e90d1)
diff --git a/src/primer/server/services/workflow_profile_service.py b/src/primer/server/services/workflow_profile_service.py
--- a/src/primer/server/services/workflow_profile_service.py
+++ b/src/primer/server/services/workflow_profile_service.py
@@ -1,5 +1,6 @@
from __future__ import annotations
+import re
from collections import Counter
from dataclasses import dataclass
@@ -25,8 +26,8 @@
"refactoring": "refactor",
"research": "investigation",
}
-_DOC_TEXT_HINTS = ("doc", "docs", "documentation", "readme", "changelog", "guide")
-_MIGRATION_TEXT_HINTS = ("migrate", "migration", "upgrade", "modernize", "deprecat", "port")
+_DOC_TEXT_RE = re.compile(r"\b(?:docs?|documentation|readme|changelog|guide)\b")
+_MIGRATION_TEXT_RE = re.compile(r"\b(?:migrat\w*|upgrade|moderniz\w*|deprecat\w*|port)\b")
@dataclass
@@ -94,8 +95,8 @@
)
verification_run_count = sum(
1
- for evidence_type in execution_types
- if evidence_type in {"test", "lint", "build", "verification"}
+ for evidence in execution_evidence
+ if _string_attr(evidence, "evidence_type") in {"test", "lint", "build", "verification"}
)
if not (
@@ -250,7 +251,7 @@
def _looks_like_docs(text: str, named_files: list[str]) -> bool:
- if any(hint in text for hint in _DOC_TEXT_HINTS):
+ if _DOC_TEXT_RE.search(text):
return True
if not named_files:
return False
@@ -263,7 +264,7 @@
def _looks_like_migration(text: str, change_shape: object | None) -> bool:
- if not any(hint in text for hint in _MIGRATION_TEXT_HINTS):
+ if not _MIGRATION_TEXT_RE.search(text):
return False
return (
_int_attr(change_shape, "files_touched_count") >= 2
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
Autofix Details
Bugbot Autofix prepared a fix for the issue found in the latest run.
- ✅ Fixed: Skill tools double-classified as both delegate and integrate
- Removed 'skill' from _INTEGRATE_HINTS so Skill-prefixed tools are only classified as delegates (via classify_tool), not also as integrate tools via substring matching.
Or push these changes by commenting:
@cursor push fada5769f4
Preview (fada5769f4)
diff --git a/src/primer/server/services/workflow_patterns.py b/src/primer/server/services/workflow_patterns.py
--- a/src/primer/server/services/workflow_patterns.py
+++ b/src/primer/server/services/workflow_patterns.py
@@ -7,7 +7,7 @@
_EDIT_HINTS = ("edit", "write", "patch", "replace", "insert", "delete", "remove", "rename", "move")
_EXECUTE_HINTS = ("bash", "terminal", "exec", "command")
_DELEGATE_HINTS = ("task", "agent", "delegate", "team", "sendmessage", "send_message")
-_INTEGRATE_HINTS = ("mcp", "plugin", "skill")
+_INTEGRATE_HINTS = ("mcp", "plugin")
def infer_workflow_steps(
Owner
Author
Remove 'skill' from _INTEGRATE_HINTS to prevent Skill:-prefixed tools from matching _is_integrate_tool via substring matching. These tools are already correctly classified as delegates via classify_tool returning 'skill' in _is_delegate_tool. Applied via @cursor push command
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 2 potential issues.
Autofix Details
Bugbot Autofix prepared fixes for both issues found in the latest run.
- ✅ Fixed: Short "rg" hint causes false search tool classification
- Removed "rg" from _SEARCH_HINTS since "ripgrep" already covers the intended case and the 2-character substring falsely matches tools containing common substrings like "merge" or "target".
- ✅ Fixed: Delegation count misses Skill-prefixed tools unlike step detection
- Added classify_tool check to _is_delegation_tool in workflow_profile_service.py so it matches the logic in _is_delegate_tool from workflow_patterns.py, ensuring Skill:xxx and orchestration tools are counted in delegation_count.
Or push these changes by commenting:
@cursor push 75f092099f
Preview (75f092099f)
diff --git a/src/primer/server/services/workflow_patterns.py b/src/primer/server/services/workflow_patterns.py
--- a/src/primer/server/services/workflow_patterns.py
+++ b/src/primer/server/services/workflow_patterns.py
@@ -2,7 +2,7 @@
from primer.common.tool_classification import classify_tool
-_SEARCH_HINTS = ("grep", "glob", "search", "find", "ripgrep", "rg")
+_SEARCH_HINTS = ("grep", "glob", "search", "find", "ripgrep")
_READ_HINTS = ("read", "fetch", "open", "cat", "view")
_EDIT_HINTS = ("edit", "write", "patch", "replace", "insert", "delete", "remove", "rename", "move")
_EXECUTE_HINTS = ("bash", "terminal", "exec", "command")
diff --git a/src/primer/server/services/workflow_profile_service.py b/src/primer/server/services/workflow_profile_service.py
--- a/src/primer/server/services/workflow_profile_service.py
+++ b/src/primer/server/services/workflow_profile_service.py
@@ -4,6 +4,7 @@
from collections import Counter
from dataclasses import dataclass
+from primer.common.tool_classification import classify_tool
from primer.server.services.workflow_patterns import (
infer_workflow_steps,
workflow_fingerprint_id,
@@ -324,7 +325,7 @@
def _is_delegation_tool(tool_name: str) -> bool:
normalized = tool_name.lower()
- return any(
+ return classify_tool(tool_name) in {"orchestration", "skill"} or any(
hint in normalized
for hint in (
"task",
Owner
Author
…tion count with step detection - Remove 'rg' from _SEARCH_HINTS; 'ripgrep' already covers the intended case and the 2-char substring falsely matches tools containing 'merge', 'target', etc. - Add classify_tool check to _is_delegation_tool in workflow_profile_service so Skill:xxx and orchestration tools are counted in delegation_count, matching the logic used by _is_delegate_tool in workflow_patterns. Applied via @cursor push command
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 2 potential issues.
Autofix Details
Bugbot Autofix prepared fixes for both issues found in the latest run.
- ✅ Fixed: Migration regex matches common "port" network term
- Changed regex from
\bport\bto\bport(?:ing|ed)\bto only match verb forms relevant to code migration, avoiding false matches on the networking term "port".
- Changed regex from
- ✅ Fixed: Docs heuristic too aggressive on text-only matching
- Removed the text-only early return so that a text regex match now requires at least one documentation file in named_files to classify as docs archetype.
Or push these changes by commenting:
@cursor push c3bb06aea5
Preview (c3bb06aea5)
diff --git a/src/primer/server/services/workflow_profile_service.py b/src/primer/server/services/workflow_profile_service.py
--- a/src/primer/server/services/workflow_profile_service.py
+++ b/src/primer/server/services/workflow_profile_service.py
@@ -28,7 +28,7 @@
"research": "investigation",
}
_DOC_TEXT_RE = re.compile(r"\b(?:docs?|documentation|readme|changelog|guide)\b")
-_MIGRATION_TEXT_RE = re.compile(r"\b(?:migrat\w*|upgrade|moderniz\w*|deprecat\w*|port)\b")
+_MIGRATION_TEXT_RE = re.compile(r"\b(?:migrat\w*|upgrade|moderniz\w*|deprecat\w*|port(?:ing|ed))\b")
@dataclass
@@ -252,8 +252,7 @@
def _looks_like_docs(text: str, named_files: list[str]) -> bool:
- if _DOC_TEXT_RE.search(text):
- return True
+ text_match = bool(_DOC_TEXT_RE.search(text))
if not named_files:
return False
doc_files = 0
@@ -261,6 +260,8 @@
normalized = path.lower()
if normalized.endswith((".md", ".mdx", ".rst", ".txt")) or "/docs/" in normalized:
doc_files += 1
+ if text_match and doc_files > 0:
+ return True
return doc_files > 0 and doc_files >= max(1, len(named_files) // 2)| "research": "investigation", | ||
| } | ||
| _DOC_TEXT_RE = re.compile(r"\b(?:docs?|documentation|readme|changelog|guide)\b") | ||
| _MIGRATION_TEXT_RE = re.compile(r"\b(?:migrat\w*|upgrade|moderniz\w*|deprecat\w*|port)\b") |
There was a problem hiding this comment.
Migration regex matches common "port" network term
Medium Severity
_MIGRATION_TEXT_RE includes \bport\b which matches the standalone word "port" — an extremely common technical term for network ports (e.g., "listen on port 3000", "bind to port 8080"). Combined with the weak secondary check in _looks_like_migration requiring only files_touched_count >= 2, this causes sessions about server configuration or networking to be falsely classified as migration archetype. Since migration is checked before debugging and feature delivery in _infer_archetype, affected sessions get the wrong archetype.
Additional Locations (1)
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 2 potential issues.
There are 3 total unresolved issues (including 1 from previous review).
Bugbot Autofix prepared fixes for both issues found in the latest run.
- ✅ Fixed: Duplicated delegation tool classification function across files
- Removed the duplicate
_is_delegation_toolfromworkflow_profile_service.pyand reused the canonicalis_delegate_tool(renamed from_is_delegate_tool) imported fromworkflow_patterns.py.
- Removed the duplicate
- ✅ Fixed: Overly broad
.txtextension in docs file detection- Removed
.txtfrom the doc file extension check in_looks_like_docs, keeping only reliably documentation-specific extensions (.md,.mdx,.rst) to prevent false positives on files likerequirements.txt.
- Removed
Or push these changes by commenting:
@cursor push 7d768e86a8
Preview (7d768e86a8)
diff --git a/src/primer/server/services/workflow_patterns.py b/src/primer/server/services/workflow_patterns.py
--- a/src/primer/server/services/workflow_patterns.py
+++ b/src/primer/server/services/workflow_patterns.py
@@ -32,7 +32,7 @@
steps.append("test")
if recovery_strategies and recovery_strategies.intersection({"edit_fix", "revert_or_reset"}):
steps.append("fix")
- if any(_is_delegate_tool(name) for name in tool_names):
+ if any(is_delegate_tool(name) for name in tool_names):
steps.append("delegate")
if any(_is_integrate_tool(name) for name in tool_names):
steps.append("integrate")
@@ -79,7 +79,7 @@
return normalized == "bash" or any(hint in normalized for hint in _EXECUTE_HINTS)
-def _is_delegate_tool(name: str) -> bool:
+def is_delegate_tool(name: str) -> bool:
normalized = _normalized_tool_name(name)
return classify_tool(name) in {"orchestration", "skill"} or any(
hint in normalized for hint in _DELEGATE_HINTS
diff --git a/src/primer/server/services/workflow_profile_service.py b/src/primer/server/services/workflow_profile_service.py
--- a/src/primer/server/services/workflow_profile_service.py
+++ b/src/primer/server/services/workflow_profile_service.py
@@ -4,9 +4,9 @@
from collections import Counter
from dataclasses import dataclass
-from primer.common.tool_classification import classify_tool
from primer.server.services.workflow_patterns import (
infer_workflow_steps,
+ is_delegate_tool,
workflow_fingerprint_id,
workflow_fingerprint_label,
)
@@ -92,7 +92,7 @@
label = workflow_fingerprint_label(fingerprint_type, steps) if fingerprint_id else None
top_tools = [tool_name for tool_name, _count in tool_counts.most_common(4)]
delegation_count = sum(
- count for tool_name, count in tool_counts.items() if _is_delegation_tool(tool_name)
+ count for tool_name, count in tool_counts.items() if is_delegate_tool(tool_name)
)
verification_run_count = sum(
1
@@ -258,7 +258,7 @@
doc_files = 0
for path in named_files:
normalized = path.lower()
- if normalized.endswith((".md", ".mdx", ".rst", ".txt")) or "/docs/" in normalized:
+ if normalized.endswith((".md", ".mdx", ".rst")) or "/docs/" in normalized:
doc_files += 1
if text_match and doc_files > 0:
return True
@@ -324,21 +324,6 @@
)
-def _is_delegation_tool(tool_name: str) -> bool:
- normalized = tool_name.lower()
- return classify_tool(tool_name) in {"orchestration", "skill"} or any(
- hint in normalized
- for hint in (
- "task",
- "agent",
- "delegate",
- "team",
- "sendmessage",
- "send_message",
- )
- )
-
-
def _bool_attr(value: object | None, field: str) -> bool:
if value is None:
return False| "sendmessage", | ||
| "send_message", | ||
| ) | ||
| ) |
There was a problem hiding this comment.
Duplicated delegation tool classification function across files
Low Severity
_is_delegation_tool in workflow_profile_service.py is functionally identical to _is_delegate_tool in workflow_patterns.py — both check classify_tool membership in {"orchestration", "skill"} and the same set of substring hints. Since workflow_profile_service.py already imports from workflow_patterns.py, this is pure duplication that risks the two implementations drifting apart.
Additional Locations (1)
| doc_files += 1 | ||
| if text_match and doc_files > 0: | ||
| return True | ||
| return doc_files > 0 and doc_files >= max(1, len(named_files) // 2) |
There was a problem hiding this comment.
Overly broad .txt extension in docs file detection
Low Severity
_looks_like_docs counts any file ending in .txt as a documentation file. Common non-doc files like requirements.txt or LICENSE.txt would trigger the docs archetype when they make up a majority of touched files — even without any text hints about documentation — because the final return path (doc_files >= max(1, len(named_files) // 2)) doesn't require text_match.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Validation
Note
Medium Risk
Adds a new persisted derived-data table and wires it into ingestion and analytics paths; mistakes could impact session ingest performance or produce incorrect workflow fingerprint/archetype analytics until backfilled.
Overview
Adds persisted workflow profiles for sessions. Introduces
session_workflow_profiles(Alembic migration + SQLAlchemy model) to store workflow fingerprint/label, step sequence, archetype (+ source/reason), top tools, delegation count, and verification run count.Derives and upserts profiles during ingest and surfaces them in product. Ingest now computes
workflow_profilevia newworkflow_profile_serviceheuristics and includes it inSessionDetailResponse; the session detail UI renders a new “Workflow Profile” card when present.Reuses stored profiles in analytics. Project workflow summary and workflow playbook generation prefer persisted profile
steps/fingerprint_id/labeland fall back to step inference when missing;infer_workflow_stepsis expanded to incorporate execution evidence, recovery strategies, and change-shape mutation signals.Written by Cursor Bugbot for commit 8ed9c3d. This will update automatically on new commits. Configure here.