find-in-circleback
Answer "what did we say, decide, or promise" from Circleback meetings — search by attendee, company, tag, or date before keyword,…
Root-cause a production error with Sentry's evidence before touching code — pull the issue, read the stack trace and breadcrumbs, and separate the crash from its trigger. Use when the user pastes a Sentry link or issue ID, reports a production error, or asks why something is
Claude Code
$ npx -y skills add paradevai/paradev --skill debug-with-sentry --agent claude-code
Codex
$ npx skills add paradevai/paradev --skill debug-with-sentry -a codex -y
Cursor
$ npx skills add paradevai/paradev --skill debug-with-sentry -a cursor -y
Gemini CLI
$ npx skills add paradevai/paradev --skill debug-with-sentry -a gemini-cli -y
GitHub Copilot
$ npx skills add paradevai/paradev --skill debug-with-sentry -a github-copilot -y
OpenCode
$ npx skills add paradevai/paradev --skill debug-with-sentry -a opencode -y
Windsurf
$ npx skills add paradevai/paradev --skill debug-with-sentry -a windsurf -y
name: debug-with-sentry description: Root-cause a production error with Sentry's evidence before touching code — pull the issue, read the stack trace and breadcrumbs, and separate the crash from its trigger. Use when the user pastes a Sentry link or issue ID, reports a production error, or asks why something is crashing for users. argument-hint: the Sentry issue URL, short ID, or a description of the error allowed-tools: mcp__sentry__get_sentry_resource, mcp__sentry__search_issues, mcp__sentry__search_errors, mcp__sentry__analyze_issue_with_seer, mcp__sentry__search_sentry_tools, mcp__sentry__execute_sentry_tool
A Sentry issue is a claim about what broke. The job is to turn it into a claim about *why* it broke, backed by the event data, before anyone edits code.
A pasted URL or short ID (`PROJECT-123`) goes straight to `get_sentry_resource` (pass the URL, or `resourceType: 'issue'` with the short ID). A described error ("users can't check out") goes through `search_issues` first, but treat the results as candidates: match on the error type and the affected code path, not on title similarity. If two issues share a message but differ in stack trace, they are two bugs.
Stack traces, breadcrumbs, tag values and releases are catalog tools: find them with `search_sentry_tools`, then call `execute_sentry_tool` with `name` set to `get_event_stacktrace`, `get_issue_breadcrumbs`, `get_issue_tag_values` or `find_releases`.
Fetch the newest event's stack trace. The top frame is where it *crashed*, not where it *broke* — walk down until you leave library code and hit the first frame owned by this repo. That frame and its local variables are the starting point. Cross-check the file against the working tree: production runs a release, and the line numbers may belong to code that has since changed. `find_releases` tells you which version the events come from.
Before proposing a cause, check what the events have in common:
release? A bug scoped to a dimension names its own trigger.
is a different bug than the same crash on first click.
release's diff. One that grew slowly points at data or traffic.
Seer (`analyze_issue_with_seer`) is worth running for a second opinion, but verify its suggested cause against the frames yourself before repeating it.
The finding is three sentences: what fails, why, and who it hits (volume, affected users, releases). Link the issue. If the user asks for a fix, fix the frame you identified — and only then consider `update_issue` to assign or resolve, never as a way to make the dashboard quiet.
threw on bad input from three frames below.
different bugs can share one, and grouping is heuristic.
evidence.
back as a new-looking issue and erases its history.
Paradev is an open-source Codex App and Conductor alternative. Orchestrate an army coding agents in parallel with your own subscriptions.
Answer "what did we say, decide, or promise" from Circleback meetings — search by attendee, company, tag, or date before keyword,…
Build a brief for an upcoming meeting from Circleback — the calendar event, the history with the same people and company, open…
Work the action items Circleback captured — find what is assigned to this user, do the ones the workspace can do, and update…
Diagnose a failing GitHub Actions run — find the first real error in the logs, tell a flake apart from a genuine failure, and…
Triage a GitHub issue into something actionable — reproduce the claim, find duplicates, judge severity, and apply labels,…
Answer "what did we say, decide, or promise" from Granola meeting notes — narrow to the meetings that can hold the answer, read…
Direct actions on one Orchestrator session or PR: stop or interrupt, resume, merge a PR, start a review, screenshot a session's…
Your capability catalog — read this at boot. Lists the temporal date-range skills and the external integrations (reached via the…
You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or…