ai-support

name	ai-support
description	Investigates customer-reported issues with structure: reproduces, traces to code, documents resolution, builds a searchable knowledge base organized by ticket ID. Trigger for 'a user is reporting that', 'customer complaint', 'support ticket', 'investigate this bug report', 'search past support cases'. Not for production incidents; use /ai-postmortem instead. Not for internal dev bugs; use /ai-debug instead. Not for feature requests; open a GitHub Issue with the enhancement label.
effort	mid
argument-hint	start [ticket-id]|find [query]

Support

Structured customer-support investigation: organize findings by ticket, link to code + PRs, build a searchable knowledge base of resolved issues. Storage: .ai-engineering/support/{YYYY-MM-DD}/{ticket-id}/investigation.md (date-organized for chronological browsing).

/ai-support start TICKET-4521    # start investigation (any ID format)
/ai-support find timeout         # search past investigations
/ai-support find                 # list all investigations

Modes

start — new investigation

1. Check existing — if a {ticket-id} directory exists under .ai-engineering/support/, resume it instead of duplicating.
2. Create — .ai-engineering/support/{date}/{ticket-id}/.
3. Scaffold investigation.md from template:

# {ticket-id}: {title}

**Date**: YYYY-MM-DD
**Customer**: {name/org if known}
**Status**: investigating | resolved | escalated
**Priority**: p1 | p2 | p3

## Issue
{Customer's description -- verbatim or summarized}

## Environment
- Product version:
- OS/Platform:
- Configuration:

## Steps to Reproduce
1. {Step}
2. {Expected vs actual behavior}

## Findings
{Root cause analysis}

## Resolution
{Fix applied, workaround, or escalation path}

## Related
- Code: {file paths}
- PR: {links}
- Notes: {links to /ai-note entries}

4. Investigate + update — explore relevant code paths, check recent changes and error patterns; keep investigation.md current as findings emerge.

find [query] — search investigations

1. Search — scan .ai-engineering/support/ for matching content.
2. Rank — recency, then relevance.
3. Present — ticket-id, date, title, status, resolution summary.

Workflow

Principles: §10.4 DRY (one investigation template + knowledge base, no per-ticket ad-hoc format).

1. Reproduce — attempt to reproduce locally using the reported steps.
2. Isolate — narrow to the specific code path, configuration, or data condition.
3. Root cause — identify why (bug, misconfiguration, edge case, expected behavior).
4. Resolve — one of: Fix (open a PR via /ai-pr, link it), Workaround (document steps), Escalate (mark escalated + reason + target team), Won't fix (document rationale).

Examples

User: "a user is reporting timeouts on TICKET-4521, investigate"

/ai-support start TICKET-4521

Scaffolds .ai-engineering/support/2026-05-08/TICKET-4521/investigation.md, attempts to reproduce, traces affected code paths, documents findings.

Integration

Called by: user directly when triaging a customer report. Calls: /ai-pr (when fixing requires a code change). See also: /ai-postmortem (production incidents), /ai-debug (internal-only bugs), /ai-note (cross-link findings).

$ARGUMENTS