jira-task
Implement a Jira ticket end to end - Haiku scouts and TDD implementers, Sonnet investigators, Opus plans and reviews.
Skill by J-h-o
Chalk Labs
Need help deploying jira-task?
Chalk Labs sets up agents like this inside real teams: connected to your tools, tested on your data, with a human in the loop where it matters.
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/j-h-o-jira-task
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: testing, productivity
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md and code. Checked Nov 1, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
- Runs as
- Node.js package
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/jira-task
What's inside
- Skill name
jira-task- Description
- Implement a Jira ticket end to end - Haiku scouts and TDD implementers, Sonnet investigators, Opus plans and reviews. Use when the user runs /jira-task or pastes a Jira ticket to implement.
- License
- MIT
- Instructions
- 65 lines in SKILL.md · view on GitHub
Before installing
- Ships executable scripts
Files (2)
- SKILL.md
- jira-execute.js
SKILL.md (first lines)
# Jira task workflow
Ticket (title, description, acceptance criteria; images may be attached to the message):
$ARGUMENTS
If no ticket content is above or attached, ask me to paste it and stop.
One ticket per session: everything from earlier work stays in this context and makes every turn more expensive. If this session already handled a ticket, recommend a new session before you start.
This skill is the whole process: don't load brainstorming, planning, or subagent-execution process skills on top of it.
The craft rules in your context apply to every phase, and subagents receive them automatically. If they are not in your context (craft mode is off), tell me and ask whether to run `/craft full` first.
## Your role: orchestrator only
You plan, delegate, verify, and decide. Subagents do all the reading, searching, debugging, browsing, and coding, because everything you do yourself stays in this large, expensive context for the rest of the run.
- A guard hook enforces this for the rest of this session. Your Read, Grep, Glob, Edit, Write, web, and browser calls are denied. Bash is limited to git and test, build and app runners, plus ordinary shell around them (loops, `echo` of exit codes, `$(...)`), with output going to temp files or filters. Trim their output (for example `| tail -40`).
- When you start a command in the background, wait for its completion notice. Don't `sleep` and then read its output; the harness blocks that.
- When you need a fact, ask a subagent a precise question and use its answer. For a follow-up, message the same subagent with SendMessage: it keeps its full context, including everything it left out of its report.
- If an investigator reports a block, ask me to unblock it, then resume that same investigator with SendMessage instead of starting a new one.
- Browser work, including checking a fix in the running app (web or mobile), always goes to an `investigator`. Tell it how to start the app and which file holds the test accounts: the path, never the values.
- If I want you to work directly, I'll end the run with `/jira-task done`.
| Need | Delegate to |
|---|---|
| Where is X, what pattern do we use, which test command | `scout` (Haiku) |
| Why does X happen, reproduce a bug, check behavior in the browser | `investigator` (Sonnet) |
| Change code for one well-specified task | `implementer` (Haiku) |
| Review the finished diff | `reviewer` (Opus) |
## 1. Understand
- Restate the ticket in 3-5 lines and list its acceptance criteria. Write criteria from the description if the ticket has none.
- Images: only you can see them, subagents can't. Transcribe what matters into words: layout, labels and copy, states, colors, error messages, and the data shown.
- Ask about real ambiguities now, in one batch. Don't ask what the code can answer.
## 2. Explore
- Dispatch `scout`s (in parallel when the questions are independent) for the relevant code, reusable helpers, test patterns, and the test, lint, and typecheck commands.
- For a bug, or any behavior you don't yet understand, dispatch an `investigator` for the root cause before you plan. Give it the symptoms, the environment details, and what "fixed" looks like.
## 3. Plan — stop for my approval
- Break the work into small tasks. Each one names the files it touches (including tests and shared files), the tasks it depends on, the behavior, the test to write first, how to verify it, and its commit message.
- Map every acceptance criterion to a task.
- Group the tasks into waves. A wave holds tasks whose dependencies are already done and whose touched files don't overlap, at most 4 per wave. Hotspots never share a wave: package manifests and lockfiles, shared types and barrel/index files, route or DI registries, migrations, i18n files, global config, and shared test setup. When in doubt, put the task in its own wave.
- Propose a branch name that follows the repo's existing convention (ask a scout) and includes the ticket key.
- Present the plan with its waves and wait for my OK.
## 4. Execute
- Create the branch, and record its base with `git rev-parse HEAD`.
- Launch the workflow once: call the Workflow tool with `scriptPath` set to the absolute path of `jira-execute.js` in this skill's base directory, and the plan as `args`, a JSON object rather than a string:
```
{ key, repoRoot, base, agentPrefix, acceptanceCriteria: [...], imageNotes, fullVerify: [...],
waves: [[{ n, title, files: [...], spec, testFirst, verify: [...], commit, criteria: [...] }]] }
```
`repoRoot` is this working tree's absolute path. `agentPrefix` is `jho-claude-kit:` when your agent types carry that prefix, otherwise empty. `imageNotes` holds the transcribed image details, `fullVerify` the full test, lint and typecheck commands, and each task's `criteria` the indexes of the acceptance criteria it serves.
- Don't poll the run; wait for its completion notice. It implements each wave test-first up the Haiku, Sonnet, Opus ladder, brings parallel waves together through detached worktrees, verifies each wave and then the whole change, and has the `reviewer` check the diff.
- When it returns, remove every path in its `worktrees` with `git worktree remove <path>` in one Bash call.
## 5. Handle the result
- `needsReplan` tasks go back to planning with me. For `needsInvestigation` tasks, dispatch an `investigator` with the reason, then re-plan them. Report `blocked` tasks with their reason and the models tried.
- Once those are resolved, launch `jira-execute.js` again with only the waves that didn't finish.
- Judge the reviewer's findings yourself. Send the valid blockers and should-fixes to one `implementer` as a batch, with this working tree as its worktree path so it commits. Run `fullVerify` and check the exit codes. Re-review only if the fixes were substantial.
## 6. Report
- What changed, mapped to the acceptance criteria.
- Verification evidence: commands and results.
- Commits made, and anything left open or worth flagging on the ticket.
- Don't push or open a PR unless I ask. Remind me that `/jira-task done` ends the guard if I want to keep working in this session.From J-h-o/jho-claude-kit, MIT.
Signals
- Stars
- 1
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Oct 14, 2026 (22 days ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/J-h-o/jho-claude-kit
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from J-h-o
Done for you
Rather not set it up yourself?
Chalk Labs scopes, builds and runs agents for your team: the task audit, the eval harness, the integrations and the weekly ops report. Fixed scope, priced before we start.
See how we build AI agents →Read the case studies
Chalk Labs is a two-founder AI and growth studio. We build AI agents, CRMs and MVPs for Web3 and AI companies, and have shipped 50+ product and token launches.
- launches shipped
- 50+
- launches shipped
- media placements
- 500+
- media placements
- founders on every account
- 2
- founders on every account
Press we've placed: Forbes · CoinDesk · Decrypt · Cointelegraph · The Block
Rahil Jain, Co-founder, Chalk Labs. Ex-Goldman Sachs engineer. You talk to the person who builds it.
Get jira-task running in your business
Tell us what you want to automate. A founder replies within one working day with a plan, not a pitch deck.
Cite this page
jira-task skill. Agent Mart. https://agentmart.dev/skills/j-h-o-jira-task (last updated 2026-10-24)