backlog-db

name	backlog-db
description	Interact with Gardener backlog SQLite storage and create new backlog tasks.

Backlog Database Skill

Use this skill when you need to inspect Gardener backlog rows or create new backlog tasks quickly.

CLI

• Canonical: cargo run -q -p gardener --bin backlog-db --
• Compatibility shim: scripts/backlog-db.sh

Common operations

• List the latest backlog rows:

  • cargo run -q -p gardener --bin backlog-db -- list
  • GARDENER_DB_PATH=PATH cargo run -q -p gardener --bin backlog-db -- list

• Create a feature task:

  • cargo run -q -p gardener --bin backlog-db -- add --title "GARD-xx: Your ticket" --details "What to fix" --priority P1 --scope runtime

• Provide overrides:

  • --kind, --status, --source, --id, --db

• Inspect or safely edit an exact row:

  • cargo run -q -p gardener --bin backlog-db -- show --id TASK_ID
  • cargo run -q -p gardener --bin backlog-db -- update --id TASK_ID --status complete --rationale "manual recovery" --clear-lease
  • cargo run -q -p gardener --bin backlog-db -- retire --id TASK_ID --status failed --rationale "duplicate"

Runbook

• cargo run -q -p gardener --bin backlog-db -- runbook prints docs/runbooks/backlog-operations.md.
• ./scripts/backlog-db.sh runbook still works via the compatibility shim.
• Use this runbook for common manual operations, state glossary, and recovery guidance.

Required fields

• --title and --details are required for add.
• priority should be one of P0, P1, P2.
• --kind must be one of: feature, maintenance, quality_gap, bugfix, infra, merge_conflict, pr_collision. Always use snake_case — the Rust parser rejects PascalCase (e.g. QualityGap will cause a database conversion error).
• --status should be one of: ready, leased, in_progress, merge_pending, complete, failed, unresolved.
• If --id is omitted, generated task id is manual:<scope>:auto-<unix_ms>.

Notes

• The Rust CLI routes writes through BacklogStore; it does not write raw SQL from the CLI layer.
• Manual CLI operations still expect an existing SQLite file.
• For large/interactive workflows, use the run command and --report options documented in README.md.