triage

Fix production issues. Run audit, investigate, fix, postmortem.

Safety Notice

This listing is imported from skills.sh public index metadata. Review upstream SKILL.md and repository scripts before running.

Copy this and send it to your AI assistant to learn

Install skill "triage" with this command: npx skills add phrazzld/claude-config/phrazzld-claude-config-triage

/triage

Fix production issues. Run audit, investigate, fix, postmortem.

This is a fixer. It uses /check-production as its primitive. Use /log-production-issues to create issues instead of fixing.

Usage

/triage # Audit and fix highest priority (default) /triage investigate VOL-456 # Deep dive on specific Sentry issue /triage investigate-ci 12345 # Deep dive on specific CI run failure /triage fix # Create PR for current fix /triage postmortem VOL-456 # Generate postmortem after merge

Stage 1: Production Audit

Command: /triage or /triage status

Invoke /check-production primitive for parallel checks:

  • Sentry - Unresolved issues via triage scripts

  • Vercel logs - Recent errors in stream

  • Health endpoints - /api/health response

  • GitHub CI/CD - Failed workflow runs

Output format:

TRIAGE STATUS - 2026-01-23 15:30

SENTRY (volume-fitness) [P0] 3 unresolved issues Top: VOL-456 "PaymentIntent failed" (Score: 147, 23 users)

GITHUB CI/CD [P1] Main branch failing: "CI" workflow (run #1234) Failed: Type check - 2h ago [P2] 2 feature branches blocked

VERCEL LOGS [OK] No errors in last 10 minutes

HEALTH ENDPOINTS [OK] volume.fitness/api/health (200, 45ms)

RECOMMENDATION:

  1. Investigate VOL-456 immediately - 23 users affected Run: /triage investigate VOL-456
  2. Fix main branch CI - blocking all deploys Run: /triage investigate-ci 1234

If all clean: "All systems nominal. No action required."

Stage 2: Investigate

Delegation Pattern

For complex issues, delegate investigation to agentic tools (see /delegate ):

  • Codex — Code archaeology, stack trace analysis, debugging

  • Gemini — Research current patterns, check for known issues

  • Thinktank — Validate proposed fix before implementing

Sentry Issues

Command: /triage investigate ISSUE-ID

Actions:

  • Fetch full issue context from Sentry

  • Create branch: fix/ISSUE-ID-description

  • Load affected files from stack trace

  • Check git history for related changes

  • Form root cause hypothesis (delegate to Codex for complex traces)

Output: Investigation summary with hypothesis and next steps.

CI/CD Failures

Command: /triage investigate-ci RUN-ID

Actions:

  • Fetch failed workflow run details gh run view RUN-ID --log-failed

  • Identify failed step and error message

  • Create branch: fix/ci-[workflow-name]-[date]

  • Load affected files based on error

  • Check recent commits that may have caused regression

Common CI failure patterns:

Failure Type Typical Cause Fix Approach

Type check New code with type errors Fix types locally, push

Lint Style violations Run pnpm lint --fix

Test Broken/flaky tests Run tests locally, fix or skip flaky

Build Missing deps, config issues Check package.json, build config

Deploy Env vars, permissions Check Vercel/platform settings

Output: CI investigation summary with specific error and fix approach.

Stage 3: Fix

Command: /triage fix

Prerequisites: On fix/ branch with changes.

Actions:

  • Run tests to verify fix

  • Create PR with standard format

  • Link Sentry issue in PR description

PR format:

Summary

[Fix description]

Sentry Issue

  • ID: ISSUE-ID
  • Users affected: N
  • First seen: DATE

Test Plan

  • Test case 1
  • Test case 2

Stage 4: Postmortem

Command: /triage postmortem ISSUE-ID

Prerequisites: Fix deployed (PR merged).

Actions:

  • Verify no new errors in Sentry

  • Generate postmortem document from template

  • Resolve Sentry issue

  • Create docs/postmortems/YYYY-MM-DD-ISSUE-ID.md

Scripts

Via Sentry MCP (Preferred)

When Sentry MCP is configured, use direct queries:

  • "Show me unresolved errors in production"

  • "What's the triage score for issue VOL-456?"

  • "Get full context for the top error"

Via CLI Scripts

Multi-source orchestrator

~/.claude/skills/triage/scripts/check_all_sources.sh

Individual checks

~/.claude/skills/triage/scripts/check_sentry.sh ~/.claude/skills/triage/scripts/check_vercel_logs.sh ~/.claude/skills/triage/scripts/check_health_endpoints.sh

Sentry CLI directly

sentry-cli issues list --project=$SENTRY_PROJECT --status=unresolved sentry-cli issues describe ISSUE-ID

Postmortem generator

~/.claude/skills/triage/scripts/generate_postmortem.sh ISSUE-ID

Via GitHub CLI

List failed runs on main branch

gh run list --branch main --status failure --limit 10

List all recent failures

gh run list --status failure --limit 10

View failed run details

gh run view RUN-ID

View only failed step logs

gh run view RUN-ID --log-failed

Re-run failed jobs (after fix pushed)

gh run rerun RUN-ID --failed

Watch a run in progress

gh run watch RUN-ID

Workflow

/triage | v [Issues found?] | +-- Sentry issue --> /triage investigate ISSUE-ID | | | v | [Fix locally] | | | v | /triage fix (creates PR) | | | v | [PR merged & deployed] | | | v | /triage postmortem ISSUE-ID | +-- CI failure --> /triage investigate-ci RUN-ID | | | v | [Fix locally, push] | | | v | [CI re-runs automatically] | | | v | [Verify CI green] | +-- No issues --> "All systems nominal"

Environment Variables

Required for Sentry

SENTRY_AUTH_TOKEN # or SENTRY_MASTER_TOKEN SENTRY_ORG # Organization slug

Auto-detected per project

SENTRY_PROJECT # From .sentryclirc or .env.local

Optional for Vercel

VERCEL_TOKEN # For vercel logs access

MCP Configuration (Recommended)

For AI-assisted triage, configure Sentry MCP:

{ "mcpServers": { "sentry": { "url": "https://mcp.sentry.dev/mcp", "transport": "http" } } }

Or local with token:

{ "mcpServers": { "sentry": { "command": "npx", "args": ["-y", "@sentry/mcp-server"], "env": { "SENTRY_AUTH_TOKEN": "your-token", "SENTRY_ORG": "your-org" } } } }

Reuses

  • ~/.claude/skills/sentry-observability/scripts/triage_score.sh

  • ~/.claude/skills/sentry-observability/scripts/issue_detail.sh

  • ~/.claude/skills/sentry-observability/scripts/resolve_issue.sh

Related

  • /check-production

  • The primitive (audit only)

  • /log-production-issues

  • Create GitHub issues from findings

  • /observability

  • Full observability setup

  • /sentry-observability

  • Sentry-specific operations

  • /verify-fix

  • Verification checklist

  • /delegate

  • Multi-AI orchestration pattern

Visual Deliverable

After completing the core workflow, generate a visual HTML summary:

  • Read ~/.claude/skills/visualize/prompts/triage-status.md

  • Read the template(s) referenced in the prompt

  • Read ~/.claude/skills/visualize/references/css-patterns.md

  • Generate self-contained HTML capturing this session's output

  • Write to ~/.agent/diagrams/triage-{date}.html

  • Open in browser: open ~/.agent/diagrams/triage-{date}.html

  • Tell the user the file path

Skip visual output if:

  • The session was trivial (single finding, quick fix)

  • The user explicitly opts out (--no-visual )

  • No browser available (SSH session)

Source Transparency

This detail page is rendered from real SKILL.md content. Trust labels are metadata-based hints, not a safety guarantee.

Open in GitHubOpen in ClawHub

Related Skills

Related by shared tags or category signals.

Security

design-audit

No summary provided by upstream source.

Repository SourceNeeds Review
-27
phrazzld
Security

changelog-audit

No summary provided by upstream source.

Repository SourceNeeds Review
-25
phrazzld
Security

billing-security

No summary provided by upstream source.

Repository SourceNeeds Review
-24
phrazzld