This project uses Beads (bd) for issue tracking.
- Start with
bd readybefore taking net-new work. - Claim a task with
bd update <id> --claim. - Create new work with
bd create --title "..." -p 2. - Use
bd show <id>for full task context andbd close <id>when the work is done. - Run
bd primewhen you need the current Beads workflow context.
TODOs.md and DONEs.md remain part of the repo during the transition:
- existing historical/open Markdown backlog stays there until explicitly migrated or retired
- new tasks created after the Beads rollout should go into
bd - keep
TODOs.md/DONEs.mdfor verification evidence and historical records while both systems coexist
If the index is missing or stale, update it:
gloggur index . --jsonIf gloggur is not on PATH (e.g., codex-cli), use:
scripts/gloggur status --json
scripts/gloggur index . --jsonUse semantic search when it materially reduces uncertainty, e.g.:
- You don’t know where a symbol is defined
- You need to locate where a concept is implemented
- You’re dealing with cross-cutting behavior across multiple modules
gloggur search "<query>" --top-k 5 --jsonIf gloggur is not on PATH:
scripts/gloggur search "<query>" --top-k 5 --jsonDo not use semantic search for:
- Editing files you already have open
- Small local refactors
- Obvious symbol locations (prefer normal text search / jump-to-definition)
Silent failures are absolutely forbidden. Fail fast, fail early, fail loud.
- Use
.venvfor running Python modules. - Check
gloggur status --jsonto confirm the index is current (orscripts/gloggur status --jsonif the CLI is not on PATH). - Cache data is stored in
.gloggur-cache; do not commit it. - If you add or rename files, re-run
gloggur index . --json. - Keep existing backlog/history in
TODOs.mdandDONEs.md, but track new tasks in Beads unless a task is explicitly part of the Markdown migration history.
For more detail, see docs/AGENT_INTEGRATION.md.
When handing of work, always include a conventional commit message for the changes