Getting Started
AI-native SDLC workflow
Run plan → build → review → ship with the coding agent you already use. Your agent reads the work from DolphinQ and reports back to it — nobody updates tickets by hand.
The loop
- Plan — turn a goal or spec into stories and dependencies (
create_stories_batch,add_dependencies_batch) - Build — the agent reads the story and logs progress as it works (
get_story,add_comment) - Review — link the PR and record what was verified (
add_story_artifact,add_commentwith VERIFIED) - Ship & track — update status, then ask about risk and blockers (
update_story,get_project_status)
Note: DolphinQ doesn't connect to GitHub or CI directly. Your agent closes the loop by calling these tools as it works.
1. Plan
Prompt
Read the refund spec in Notion and plan it in DolphinQ
as stories with components, dates, and dependencies.2. Build — let your agent pick up a story
Prompt
Pick up DLPQ-012. Read the story and its comments,
implement it, and log progress on the story as you go —
TASK for steps, FIX for bug fixes, BLOCKER if you get stuck.3. Review — link the PR and record verification
Prompt
I opened PR #42 for DLPQ-012. Attach it to the story,
post a VERIFIED comment with what you tested,
and set the story to Completed.The comment lands on the story as a formatted, tagged update — written by the agent that did the work.
4. Assess and track
Prompt
What's blocking the release? Analyze DLPQ-015 and save it.Make it automatic
Add these rules to your project's CLAUDE.md (Claude Code) or AGENTS.md (Codex) so your agent follows the loop without being asked:
CLAUDE.md / AGENTS.md
## DolphinQ workflow
- Before starting work on a story, read it with get_story (including comments).
- Log meaningful progress with add_comment: TASK for steps,
FIX for bug fixes, BLOCKER when stuck.
- When a PR is opened, attach it with add_story_artifact.
- After verifying, post a VERIFIED comment saying what was tested,
then set the status with update_story.
- Write comments in markdown; wrap code in fenced code blocks.