/migration — Create a Migration Ticket + AgDR

Read .claude/rules/writing-standard.md. Use the controlled technical writing profile for the migration ticket and AgDR. Lead with the outcome and keep required detail precise.

Migrations are high-blast-radius: data loss, downtime, lock contention, cross-service coordination. ApexYard treats them as a distinct class of change from code — a migration PR needs:

1. A tracker issue with the migration label (plus priority) that captures the plan.
2. A migration AgDR at docs/agdr/AgDR-NNNN-migration-<slug>.md that captures the options considered, rollback plan, downtime estimate, cross-service consumers, data volume, testing plan, and observability.

require-migration-ticket.sh (the matching PreToolUse hook) refuses to let you write to migration paths without both artefacts in place. This skill produces both in one pass.

Path resolution

Read the registry path via portfolio_registry, the per-project docs dir via portfolio_projects_dir, and the ideas backlog via portfolio_ideas_backlog — all from .claude/hooks/_lib-portfolio-paths.sh. Source the helper at the top of any bash block that touches those paths:

source "$(git rev-parse --show-toplevel)/.claude/hooks/_lib-read-config.sh"
source "$(git rev-parse --show-toplevel)/.claude/hooks/_lib-portfolio-paths.sh"
registry=$(portfolio_registry)

Defaults match today's single-fork layout (./apexyard.projects.yaml, ./projects, ./projects/ideas-backlog.md). Adopters in split-portfolio mode override the portfolio.{registry, projects_dir, ideas_backlog} keys in .claude/project-config.json. Don't hardcode literal apexyard.projects.yaml or projects/ paths in bash blocks — the helper resolves whichever mode the adopter is in. See docs/multi-project.md.

Usage

/migration                 # prompts for everything, creates ticket + AgDR in the current project
/migration sample-app      # explicitly target a registered project

When to invoke

• Before writing a new SQL migration file
• Before adding/modifying a DynamoDB table in backend/template.yaml
• Before adding/modifying an aws_rds_* / aws_dynamodb_table resource in Terraform
• Before editing a Prisma schema in a way that will produce a migration
• Before editing TypeORM / Alembic migration directories
• Generally: BEFORE the first Write on anything the migration-gate hook blocks

Process

Use AskUserQuestion for every operator option menu in this skill. Follow .claude/rules/reporting-style.md § Operator choices. Preserve multiple selections where the menu permits them. Split menus with more than four options into sequential wizard questions. Keep single yes/no and ticket confirmation prompts as written. The prose menus below are fallbacks only when the harness lacks AskUserQuestion.

0. Write the active-issue-skill marker (REQUIRED — me2resh/apexyard#268)

Before any gh issue create (or other tracker CLI), write this skill's name to the active-issue-skill marker so require-skill-for-issue-create.sh lets the command through. At skill entry:

# Ops root through the session pin (docs/skills/shared/ops-root.md).
ops_root=$(p="${APEXYARD_OPS_PIN_DIR:-$HOME/.claude/apexyard}/ops-root-${CLAUDE_CODE_SESSION_ID:-}"; r=""; [ -z "${APEXYARD_OPS_DISABLE_PIN:-}" ] && [ -f "$p" ] && IFS= read -r r < "$p"; [ -f "$r/.claude/hooks/_lib-ops-root.sh" ] || { r="$PWD"; while [ -n "$r" ] && [ ! -f "$r/.claude/hooks/_lib-ops-root.sh" ]; do r=${r%/*}; done; }; [ -f "$r/.claude/hooks/_lib-ops-root.sh" ] && . "$r/.claude/hooks/_lib-ops-root.sh" && resolve_ops_root) && [ -n "$ops_root" ] || { echo "No ops root. See docs/skills/shared/ops-root.md." >&2; exit 1; }
mkdir -p "$ops_root/.claude/session"
marker_home="$ops_root"
. "$marker_home/.claude/hooks/_lib-session-markers.sh"
skill_marker=$(active_issue_skill_marker_path "$marker_home") || { echo "A session id is required. Restart through the harness adapter." >&2; exit 2; }
echo "migration" > "$skill_marker"

Remove the marker on every exit path (success, early-exit, user cancel, error):

marker_home="$ops_root"
. "$marker_home/.claude/hooks/_lib-session-markers.sh"
skill_marker=$(active_issue_skill_marker_path "$marker_home") || { echo "A session id is required. Restart through the harness adapter." >&2; exit 2; }
rm -f "$skill_marker"

The clear-issue-skill-marker.sh SessionStart hook clears only the current session's marker, but a clean exit should never leave one behind. See AgDR-0030.

1. Resolve the target project

If the user passed <project>, use that. Otherwise:

• If cwd is under <ops_root>/workspace/<project>/, infer project from the path
• Otherwise ask explicitly: "Which project is this migration for?"

Look up the project in apexyard.projects.yaml to get repo (the tracker) and workspace (for computing the AgDR path).

If the project isn't registered, stop — file one via /handover first, or pass the ticket as a plain ops-level change.

2. Gather the migration facts (conversational)

Ask each of the following. Each answer feeds both the issue body and the AgDR — the skill writes them into both so the user never retypes. Use AskUserQuestion for the fields with listed choices. For a choice field such as migration type, recommend the best supported value first. For a factual field such as "Rollback tested against" or "Estimated downtime", do not recommend a value the operator has not confirmed. Put the most conservative value first (for example not tested), and recommend another value only when the conversation already holds evidence for it. Use the listed values as a numbered prose fallback only when the harness lacks the tool.

1. One-line summary — goes in the ticket title: [Migration] <type>: <summary>
2. Migration type — schema | data | sql | orm (pick one; if it straddles, use the most invasive)
3. Affected tables / entities — comma-separated list. Required non-empty.
4. Rollback plan — free text, required non-empty. Ask "if this goes wrong in prod at 3am, what exactly does the rollback runbook look like?" Capture the actual steps, not a promise to figure it out later.
5. Rollback tested against — staging | copy of prod | unit fixture | not tested. If "not tested", flag it in the AgDR and tell the user this is a blocker for prod apply (not for creating the ticket).
6. Estimated downtime — none | seconds | minutes | hours. Plus reasoning: "why this much / this little?"
7. Cross-service consumers — list every service that reads or writes the affected tables. If genuinely none, say so (makes review easier).
8. Deploy-order constraint — if any service must deploy before or after this migration, record it.
9. Data volume — rough row/item count. If unknown, note that and add "size check" to the testing plan.
10. Testing plan — dev smoke command, staging verify steps, canary / phased rollout if applicable.
11. Observability — metrics/logs that will confirm success during and after. If the project already has dashboards, link them.
12. Priority — P0 | P1 | P2 | P3. Default P1 for migrations (blast radius).

Re-prompt if rollback plan is empty, affected tables is empty, or priority is missing — these are the three fields with no safe default.

3. Preview

Before creating anything, echo back a structured preview so the user can correct mistakes:

Ticket:
  Tracker:      <owner/repo>
  Title:        [Migration] <type>: <summary>
  Labels:       migration, <priority>
  Body:         (formatted — see below)

AgDR:
  Path:         docs/agdr/AgDR-NNNN-migration-<slug>.md
  Next number:  NNNN = max(existing AgDR ids in that dir) + 1

Ask "Create ticket + AgDR? [y/N]". Only proceed on explicit y.

4. Create the AgDR first (local write, reversible)

Resolve the migration AgDR template via the portfolio helper so adopter overrides win when present:

source "$(git rev-parse --show-toplevel)/.claude/hooks/_lib-read-config.sh"
source "$(git rev-parse --show-toplevel)/.claude/hooks/_lib-portfolio-paths.sh"
template=$(portfolio_resolve_template agdr-migration.md)   # → custom-templates/agdr-migration.md if present
cp "$template" "$resolved_agdr_path"

Single-fork adopters (no portfolio block) and adopters with no override fall straight through to templates/agdr-migration.md. Adopters who want a customised migration-AgDR shape drop their version at <private_repo>/custom-templates/agdr-migration.md. See templates/README.md for the path-mirroring convention.

Fill in:

• Frontmatter (id, timestamp, agent, model, trigger: user-prompt, status: draft, ticket — left as TBD until step 5 creates the issue and we know the number)
• The title, one-sentence summary, and every Section (Context, Options, Decision, Rollback Plan, Cross-Service Consumers, Testing Plan, Observability, Consequences) with the user's answers

AgDRs are written in the RIGHT repo:

Where the migration runs	AgDR path
Inside a managed project's own code repo	workspace/<project>/docs/agdr/AgDR-NNNN-migration-<slug>.md
Against the apexyard framework itself (rare — only for framework-level data/config migrations)	docs/agdr/AgDR-NNNN-migration-<slug>.md in the ops fork

For the NNNN id: scan the target docs/agdr/ directory for existing AgDR-\d+-.*\.md files, take the max id, increment. Zero-pad to 4 digits.

5. Create the GitHub issue

Title: [Migration] <type>: <summary>

Labels: migration (or whatever the project configures as its migration label — check .claude/project-config.json key migration_label, default migration), plus the priority label (P0 / P1 / P2 / P3).

Body — resolve the migration ticket body template via the portfolio helper so adopter overrides win when present:

source "$(git rev-parse --show-toplevel)/.claude/hooks/_lib-read-config.sh"
source "$(git rev-parse --show-toplevel)/.claude/hooks/_lib-portfolio-paths.sh"
ticket_template=$(portfolio_resolve_template tickets/migration.md)   # → custom-templates/tickets/migration.md if present

Single-fork adopters (no portfolio block) and adopters with no override fall straight through to templates/tickets/migration.md. Adopters who want a customised migration-ticket shape drop their version at <private_repo>/custom-templates/tickets/migration.md. See templates/README.md for the path-mirroring convention.

Backward-compat fallback: if portfolio_resolve_template returns empty (template file missing — partial adopter setup), fall back to the inline heredoc body below and print a one-line WARN on stderr (WARN: tickets/migration.md template missing — using inline fallback).

Body (CommonMark, must include a ref to the AgDR so require-migration-ticket.sh can verify it) — the default templates/tickets/migration.md shape is reproduced below:

## Migration

**Type**: <type>
**Affected tables/entities**: <list>
**Estimated downtime**: <level> — <reasoning>
**Data volume**: <count or "unknown">
**Priority**: <priority>

## Rollback Plan

<rollback plan verbatim>

**Tested against**: <staging | copy of prod | unit fixture | not tested>

## Cross-Service Consumers

<list or "none">

**Deploy-order constraint**: <constraint or "none">

## Testing Plan

- Dev smoke: <command>
- Staging verify: <steps>
- Canary / phased rollout: <plan or "n/a">

## Observability

<what to watch during + after>

## Agent Decision Record

Migration AgDR: `<relative-path-to-AgDR>`

## Glossary

| Term | Definition |
|------|------------|
| <term> | <definition> |

---

Created by `/migration`. Do not edit the labels off this issue — the
migration-gate hook (`require-migration-ticket.sh`) verifies the
`migration` label is present before allowing edits to migration files.

Create via the tracker abstraction (tracker_create, #670 / AgDR-0072, extended by the #709 creator sweep) so the ticket lands in this project's tracker — GitHub, GitLab, or a custom CLI — per its tracker: block in apexyard.projects.yaml. For a GitHub adopter this runs gh issue create exactly as before.

# Resolve the tracker lib (it lives in the ops fork's hooks dir) by walking up
# from the cwd; source it.
tracker_lib="$(r="$PWD"; while [ -n "$r" ] && [ "$r" != / ]; do \
  [ -f "$r/.claude/hooks/_lib-tracker.sh" ] && { echo "$r/.claude/hooks/_lib-tracker.sh"; break; }; \
  r="${r%/*}"; done)"
# shellcheck source=/dev/null
. "$tracker_lib"

# $BODY was assembled above; hand it to tracker_create via a file (arbitrary
# markdown — never inline-interpolated). Labels pass as a comma-separated list.
body_file="$(mktemp)"
printf '%s\n' "$BODY" > "$body_file"

# tracker_create <owner/repo> <title> <body_file> [<labels_csv>] → {"ref","url"}.
result="$(tracker_create "<owner/repo>" "[Migration] <type>: <summary>" "$body_file" "migration,<priority>")"
rc=$?
rm -f "$body_file"
if [ "$rc" -eq 3 ]; then
  echo "Tracker is 'none' (shape-only) — nothing was created in a tracker." >&2
  echo "File this in your external system (e.g. Jira/Linear via MCP):" >&2
  printf '%s\n' "$result"
  exit 0
elif [ "$rc" -ne 0 ] || [ -z "$result" ]; then
  echo "Migration ticket creation failed — check the tracker CLI / auth. Nothing was created." >&2
  exit 1
fi

# Capture the reference + URL for the AgDR back-fill below.
ref="$(printf '%s' "$result" | jq -r '.ref')"
url="$(printf '%s' "$result" | jq -r '.url')"

The migration label is what require-migration-ticket.sh reads to gate edits to migration files (Gate 3a). On GitHub it must already exist on the repo; on GitLab it is auto-created when first applied.

6. Back-fill the AgDR with the issue reference

Open the AgDR written in step 4 and update (using the ref / url captured above):

• Frontmatter ticket: <owner/repo>#${ref}
• The Artifacts section: add the ticket ${url}

This lets a future reader land on the AgDR and find the ticket, and land on the ticket and find the AgDR.

7. Return a summary

Single-line output (using the ref / url captured above):

Migration ticket: ${url}
Migration AgDR:   <relative-path-to-AgDR>
Next step:        run /start-ticket <owner/repo>#${ref}, then begin editing the migration files

Do NOT automatically run /start-ticket — the user may have other context to set first, and the explicit handoff makes the workflow legible. The migration gate will block edits until the marker points at this ticket.

Rules

1. Never ship without a rollback plan. If the user cannot articulate rollback steps, the migration isn't ready — the skill refuses to create the ticket until they type something in that field.
2. Never auto-assign priority. Migrations span from "trivial schema rename" (P3) to "primary-key column type change on a 100M-row table at peak traffic" (P0). Ask.
3. Never skip the AgDR. Even small migrations get one. If it feels like overkill for a 3-line change, the AgDR entries will be short — that's fine. The value is in the forcing function of thinking through rollback + observability, not in the document length.
4. Never create the AgDR under .claude/ or docs/ unless the migration IS against apexyard itself. For managed projects, the AgDR lives inside that project's repo (workspace/<project>/docs/agdr/), not in the ops fork.
5. Write AgDR first, ticket second, back-fill third — the AgDR is a local file, reversible with rm. The ticket is remote state. If anything errors after the ticket is created, the user has a ghost ticket to clean up; minimise that exposure.
6. Tell the user the next step is /start-ticket — the migration-gate hook checks the active ticket has the migration label, so they need to declare it before touching migration files.

Relation to the migration gate

Hook	require-migration-ticket.sh (fires PreToolUse on Write/Edit to migration paths)
Gate 1	Active ticket exists (same as require-active-ticket.sh)
Gate 2	The active ticket on GitHub has the migration label
Gate 3	The active ticket's body references an AgDR at docs/agdr/AgDR-\d+-.*migration.*\.md
Fail message	Points at this skill with the exact invocation to run

The skill and the gate are two halves of the same mechanism: gate detects the situation and blocks, skill builds the artefacts needed to unblock.

---

Part of ApexYard — multi-project SDLC framework for Claude Code · MIT.