x-bug-triage-plugin

Optional Team Review Flow — X Bug Triage Plugin

Overview

The primary interface is the Claude Code terminal. Users run /x-bug-triage, see results, and type review commands directly. No additional setup needed.

For team workflows, the optional claude-code-slack-channel plugin adds async Slack delivery so multiple reviewers can triage from a shared channel.

Terminal-First Flow (Default)

User runs /x-bug-triage @account
  → Claude fetches, analyzes, clusters
  → Results displayed as markdown in terminal
  → User types commands: "details 1", "file 2", "dismiss 3 noise"
  → Claude processes command, displays response
  → "confirm file 2" gates all issue creation

Optional Slack Flow (When Plugin Installed)

User runs /x-bug-triage @account
  → Results displayed in terminal (same as above)
  → ALSO sent to configured Slack channel via bridge's reply tool
  → Team members can send commands from Slack
  → Claude processes commands from both terminal and Slack

Key distinction: The triage server’s parse_review_command tool handles command parsing. Claude formats all output directly as markdown. All Slack transport goes through the separate claude-code-slack-channel plugin.

Summary Format

X Bug Triage — Run {date} {time} UTC
   Account: @{account} · Window: last {window} · {count} posts ingested

--- {n} clusters ({new} new, {existing} existing) ---

{severity_icon} {#} · {bug_signature}
     {report_count} reports · {severity} severity · {status_note}
     Owner: {team}
     Top evidence: {highest_tier_description} (Tier {n})

--- Commands ---
details <#>  ·  file <#>  ·  dismiss <#>  ·  merge <#> <issue>
escalate <#>  ·  monitor <#>  ·  snooze <#> <duration>
split <#>  ·  reroute <#>  ·  full-report

Max 20 lines for <=5 clusters
Top 5 by severity for 6+ clusters, then “N more — type full-report”

Detail View

Full cluster detail includes:

Family, surface, feature area
Report count, confidence
Severity + rationale (always show why for high/critical)
Status and time range
Evidence by tier (all tiers shown)
3 representative posts (highest quality, most distinct, most recent)
Repo evidence summary
Routing with ranked assignees
Available actions

11 Review Commands

Command	Action
`details <#>`	Show full cluster detail
`file <#>`	Generate issue draft for review
`dismiss <#> <reason>`	Suppress cluster with reason
`merge <#> <issue>`	Link cluster to existing issue
`escalate <#>`	Escalate to higher severity
`monitor <#>`	Set cluster to monitoring state
`snooze <#> <duration>`	Temporarily suppress (e.g., `snooze 3 24h`)
`split <#>`	Split cluster into sub-clusters
`reroute <#>`	Change routing recommendation
`full-report`	Show all clusters
`confirm file <#>`	Actually file the issue (after reviewing draft)

Error Handling

Situation	Response
Missing cluster number	“Which cluster?”
Invalid number	“No cluster N. Available: 1, 2, 3”
Unrecognized command	“Available commands: …”
Already filed	“Cluster N was already filed as ISSUE-XXX. Want to update instead?”

Formatting Rules

Severity icons: red_circle critical/high, yellow_circle medium, green_circle low
Initial summary: highest evidence tier only, team only
Detail view: all tiers, ranked assignees, full rationale
3 representative posts per cluster (by quality, distinctness, recency)
50 reports: show count + top 3 only
Tone: concise, factual, no hype, no exclamation marks

Degradation (Slack Only)

Slack plugin not installed → terminal-only, not an error
Slack API error → retry 3x with backoff, then save to data/reports/ as JSON
Channel not found → alert via fallback channel or log
Message too long → truncate to top 5 by severity, link to full report

This site is open source. Improve this page.