Help us improve
Share bugs, ideas, or general feedback.
From faultline
Investigates Rails production errors via the Faultline MCP server. Fetches error groups, full occurrences with backtraces and locals, and APM traces to diagnose exceptions and slow endpoints.
npx claudepluginhub dlt/faultline --plugin faultlineHow this skill is triggered — by the user, by Claude, or both
Slash command
/faultline:debuggingThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
When the user wants to debug a production error, diagnose it using the Faultline MCP tools — don't ask them to paste a stack trace from the dashboard.
Investigates a Faultline error group by fetching group details, the most recent occurrence with full backtrace and local variables, and optionally correlated traces for performance issues. Proposes a fix grounded in the data.
Implements error tracking systems, configures alerts and dashboards, sets up structured logging and error grouping for production issue resolution. Useful for monitoring and observability tasks.
Detects and diagnoses application failures by analyzing error logs, stack traces, and system state. Useful when debugging crashes or production incidents.
Share bugs, ideas, or general feedback.
When the user wants to debug a production error, diagnose it using the Faultline MCP tools — don't ask them to paste a stack trace from the dashboard.
Follow these steps; skip any that don't apply.
list_error_groups with search: and pick the most plausible match. If they gave a numeric id, skip to step 2.get_error_group with the id. The response includes a summary of the 10 most recent occurrences (no backtraces or locals — those are intentionally omitted to keep the response small).get_occurrence with the most recent occurrence id to get the full backtrace, captured local variables, filtered request params, request headers, and source context. This is the step that actually lets you fix the bug — don't propose a fix without seeing the locals.list_traces with the relevant endpoint, then get_trace for span detail.Read-only:
list_error_groups — recent groups, filter by status/since/searchget_error_group — single group + recent occurrence summaryget_occurrence — full occurrence with backtrace, locals, request datarecent_occurrences — paginate occurrences for a grouperror_stats — time-bucketed countslist_traces / get_trace — APM data (only when enable_apm is set on the host app)Mutating (only available when the Faultline app sets mcp_readonly: false):
resolve_error_group — mark resolvedignore_error_group — mark ignoredcreate_github_issue — open a GitHub issue with full contexterror: "Tool disabled: mcp_readonly is true". That's expected on production-default installs. If the user wants you to resolve/ignore/file issues directly, tell them to set c.mcp_readonly = false in their Faultline initializer.Rails.application.config.filter_parameters plus config.sanitize_fields. If you see [FILTERED] in request_params, that's by design — don't ask the user for the underlying value.