parallel-task · jdrhyne — Claude Code skill: what it does, how to install
Parses a markdown plan into sprint tasks and hands unblocked ones to parallel subagents. Fires on the explicit /parallel-task command.
MIT ★ 239 updated 2026-08-30
npx skills add jdrhyne/agent-skills --skill parallel-task -a claude-codeWhat parallel-task does
The skill reads a given plan file, locates a sprint or phase section (like ## Sprint 1:) and pulls out each task's description, file paths, acceptance criteria and validation steps. Every task gets dispatched to its own subagent with a fixed prompt template covering plan context, dependencies, criteria and instructions to report back modified files.
Tasks only run in parallel if they don't write to the same files; when scopes overlap, the skill holds the task until its dependency clears. Once a batch finishes, it marks those tasks done, logs the work, then rereads the whole plan to find newly unblocked tasks before launching the next batch. This loop continues until the plan is finished or a blocker shows up, at which point that branch of work stops and the blocker gets surfaced separately.
After all tasks are done, the skill checks the final state against the plan and runs available validation before producing a summary of files touched, changes made and what was validated or deferred.
What it helps with
- split a markdown sprint plan into discrete tasks for agents
- run several tasks at once without file-write conflicts
- detect which tasks become unblocked once others finish
- keep a work log of completed tasks with files and errors
- verify the final result against the plan's acceptance criteria
Example prompts for the agent
- /parallel-task plan.md
- /parallel-task ./plans/auth-plan.md sprint 2
- /parallel-task user-profile-plan.md phase 1
How to install parallel-task
1Skills CLI (Vercel)
npx skills add jdrhyne/agent-skills --skill parallel-task -a claude-code2Manually: put the files in place
~/.claude/skills/parallel-task/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/parallel-task/
1Skills CLI (Vercel)
npx skills add jdrhyne/agent-skills --skill parallel-task -a codex2Manually: put the files in place
.agents/skills/parallel-task/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add jdrhyne/agent-skills --skill parallel-task -a cursor2Manually: put the files in place
.cursor/skills/parallel-task/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install jdrhyne/agent-skills skills/parallel-task2Manually: put the files in place
.github/skills/parallel-task/SKILL.md1Skills CLI (Vercel)
npx skills add jdrhyne/agent-skills --skill parallel-task -a gemini-cli2Platform CLI
gemini skills install https://github.com/jdrhyne/agent-skills --path skills/parallel-taskManually: put the files in place
.devin/skills/parallel-task/SKILL.mdThe legacy .windsurf/skills/ path still works
Claude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about parallel-task
How do I install parallel-task in Claude Code?
Run npx skills add jdrhyne/agent-skills --skill parallel-task -a claude-code; the skill lands at ~/.claude/skills/parallel-task/SKILL.md.
What does the parallel-task skill do?
It reads a markdown plan broken into sprints, then launches parallel subagents on tasks that aren't blocked, holding back any that share write scope on the same files. After each batch it rereads the plan to find newly unblocked work.
When does it run subagents in parallel versus one at a time?
Tasks launch together only when they don't touch overlapping files; if a dependency between tasks isn't resolved yet, the skill waits instead of running them side by side.
Source text
SKILL.md · MIT
--- name: parallel-task description: Coordinate plan files by launching multiple parallel subagents for unblocked tasks. Triggers on explicit "/parallel-task" commands. credits: | Original skill by [@am-will](https://github.com/am-will) --- # Parallel Task Runner Use Orchestration Mode. Parse plan files and delegate tasks to parallel subagents. ## Safety Boundaries - Do not launch subagents for tasks that share write scope until the dependency is resolved. - Do not continue past blockers silently; surface the blocked task and stop that branch of work. - Do not mark tasks complete until validation from the plan has actually been checked. - Do not run this skill unless the user explicitly invoked `/parallel-task` or clearly asked for parallel agent work. ## Process ### Step 1: Parse Request Extract from user request: 1. **Plan file**: The markdown plan to read 2. **Sprint/Phase** (optional): Which section to run If sprint not provided, ask user which to run. ### Step 2: Read & Parse Plan 1. Find sprint/phase section (e.g., `## Sprint 1:`) 2. Extract task subsections (e.g., `### Task 1.1:`) 3. For each task, extract: - Task ID and name - Full content (description, location, acceptance criteria, validation) 4. Build task list ### Step 3: Launch Subagents For each task, launch subagent with: - **description**: "Implement task [ID]: [name]" - **prompt**: Use template below If there are multiple unblocked tasks, launch multiple subagents in parallel. ### Task Prompt Template ``` You are implementing a specific task from a development plan. ## Context - Plan: [filename] - [sprint/phase] - Goals: [relevant overview from plan] - Dependencies: [prerequisites for this task] - Related tasks: [tasks in same sprint] - Constraints: [risks from plan] ## Your Task **Task [ID]: [Name]** Location: [File paths] Description: [Full description] Acceptance Criteria: [List from plan] Validation: [Tests or verification from plan] ## Instructions 1. Examine and explore the plan, all relevant files & dependencies. Ensure ambiguities are cleared up. 2. Implement changes for all acceptance criteria 3. Keep work **atomic and committable** 4. For each file: read first, edit carefully, preserve formatting 5. Run validation if feasible 6. Return summary of: - Files modified/created - Changes made - How criteria are satisfied - Validation performed or deferred ## Important - Be careful with paths - Stop and describe blockers if encountered - Focus on this specific task ``` ### Step 4: Monitor & Log After subagents complete: 1. Collect results 2. **ALWAYS mark completed tasks** and update with: - Concise work log - Files modified/created - Errors or gotchas encountered ### Step 5: Repeat Now, pause and read the whole plan again to see what new tasks are unblocked. Start back at step 3. Continue launching agents on unblocked tasks in parallel. Repeat this loop until the plan is done. Do not stop until the plan is fully completed or you encounter a blocker. ## Step 6: Validate After completing all tasks, validate the plan's success by checking the final state against the original plan. Perform available tests and ensure success. If there are any issues, address them before marking the plan as complete. After validation, provide a summary of: - Files modified/created - Changes made - How criteria are satisfied - Validation performed or deferred ## Error Handling - Sprint not found: List available sprints/phases - Parse failure: Show what was tried, ask for clarification ## Example Usage ``` /parallel-task plan.md /parallel-task ./plans/auth-plan.md sprint 2 /parallel-task user-profile-plan.md phase 1 ``` ## Run Summary Template ```markdown # Sprint/Phase Run Summary ## Tasks Assigned: [N] ### Completed - Task [ID]: [Name] - [Brief summary] ### Issues - Task [ID]: [Name] - Issue: [What went wrong] - Resolution: [How resolved or what's needed] ### Blocked - Task [ID]: [Name] - Blocker: [What's preventing completion] - Next Steps: [What needs to happen] ## Overall Status [Completion summary] ## Files Modified [List of changed files] ## Next Steps [Recommendations] ```
© jdrhyne · MIT · jdrhyne/agent-skills
Similar skills
All skills for Agent setup: memory, context, orchestration →parallel-task by other authors comes first.