| name | brain-dump-workflow |
| description | Use Brain Dump entirely through the brain-dump CLI. Start ticket work, implement changes, run validation, complete work, review findings, and generate a demo without MCP. |
Brain Dump Workflow
Use the brain-dump CLI directly. Do not rely on MCP. Prefer CLI commands plus pi's built-in bash, read, edit, and write tools.
Core rule
When working on a Brain Dump ticket, follow this sequence:
- Inspect project/ticket context
- Start work with the CLI
- Implement and verify
- Complete work with a summary
- Review and log findings with the CLI
- Generate demo steps for human review
1) Inspect context
Use these commands first:
brain-dump doctor
brain-dump status --pretty
brain-dump ticket list --pretty
brain-dump context --ticket <ticket-id> --pretty
If brain-dump is not in PATH, use:
pnpm brain-dump <command>
2) Start work
Before changing code, run:
brain-dump workflow start-work --ticket <ticket-id> --pretty
This should create/check out the branch and move the ticket to in_progress.
3) Implement and verify
Use pi tools to inspect and edit files. Keep changes focused.
Discover and run this project's validation commands from docs/config. Check AGENTS.md,
CLAUDE.md, README, CONTRIBUTING, package scripts, pyproject.toml, go.mod,
Makefile/Justfile, and CI files. Do not assume pnpm, npm, TypeScript, lint, or test
scripts exist.
Examples only: package script check/test/lint, pytest/ruff when configured,
go test ./..., cargo test, dotnet test, mvn test, ./gradlew test.
If no automated validation command is discoverable, run a targeted manual smoke check
and record that no project validation command was found.
Good commit format:
git commit -m "feat(<ticket-id>): <description>"
You can also sync or link git work manually if needed:
brain-dump git sync --pretty
brain-dump git link-commit --ticket <ticket-id> --hash $(git rev-parse HEAD) --pretty
brain-dump git link-pr --ticket <ticket-id> --pr <number> --pretty
4) Complete implementation
After implementation and validation, run:
brain-dump workflow complete-work --ticket <ticket-id> --summary "<what changed>" --pretty
This should move the ticket to ai_review.
5) Review with CLI logging
Perform self-review by reading diffs and changed files.
Useful commands:
git diff --stat
git diff
brain-dump review check-complete --ticket <ticket-id> --pretty
If you find issues, log them explicitly:
brain-dump review submit-finding \
--ticket <ticket-id> \
--agent code-reviewer \
--severity major \
--category error-handling \
--description "Describe the issue" \
--file src/example.ts \
--line 42 \
--fix "Describe the intended fix" \
--pretty
After fixing an issue, mark it:
brain-dump review mark-fixed --finding <finding-id> --status fixed --description "Fixed by ..." --pretty
Re-run checks until Brain Dump reports review is complete:
brain-dump review check-complete --ticket <ticket-id> --pretty
6) Generate demo for human review
Create a JSON file with 3-7 demo steps, for example at .pi/tmp/demo-steps.json:
[
{
"order": 1,
"description": "Open the updated screen",
"expectedOutcome": "The page loads without errors",
"type": "manual"
}
]
Then run:
brain-dump review generate-demo --ticket <ticket-id> --steps-file .pi/tmp/demo-steps.json --pretty
At that point, stop and wait for human review.
Useful CLI shortcuts
brain-dump status --pretty
brain-dump log --pretty
brain-dump ticket get --ticket <ticket-id> --pretty
brain-dump review get-demo --ticket <ticket-id> --pretty
CLI parity notes
Most MCP-style operations have CLI parity, sometimes under a different resource name:
- workflow
link-commit -> brain-dump git link-commit
- workflow
link-pr -> brain-dump git link-pr
- workflow
sync-links -> brain-dump git sync
- session task actions ->
brain-dump tasks save|get|clear|snapshots
- project
find-by-path -> brain-dump project find
- admin settings actions ->
brain-dump settings get|update
- admin conversation logging actions ->
brain-dump compliance start|log|end|list|export|archive
Notes
- Prefer Brain Dump CLI over handwritten status tracking.
- Prefer
brain-dump context --ticket <ticket-id> --pretty before implementation.
- If a command fails, read the error and recover through Brain Dump CLI rather than bypassing workflow state.