- Home
- Skills
- Am Will
- Codex Skills
- Parallel Task
parallel-task_skill
- Python
448
GitHub Stars
1
Bundled Files
2 months ago
Catalog Refreshed
4 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill am-will/codex-skills --skill parallel-task- SKILL.md4.4 KB
Overview
This skill orchestrates parallel subagents to execute plan files when explicitly invoked with the /parallel-task command. It parses markdown plan documents, launches unblocked tasks in waves, monitors completion, and updates the plan with logs until all tasks are complete. The goal is reliable, ordered parallel execution while keeping plan documentation and commits consistent.
How this skill works
On invocation it extracts the target plan and optional task subset, parses task sections and dependency metadata, then builds a dependency graph. It launches all unblocked tasks in parallel using a standardized task prompt, waits for each wave to finish, validates outputs and that the plan file was updated with a work log, then repeats until every required task is complete. It enforces commit-only workflows per task and retries or escalates failed tasks.
When to use it
- Run full plans or selected task subsets from markdown plan files using /parallel-task
- Coordinate multiple implementers or subagents that must respect task dependencies
- Automate staged feature delivery where tasks must be validated and logged in-plan
- Speed up execution by running independent tasks in parallel
- Enforce consistent task completion logs and per-task commits
Best practices
- Provide a clear, well-structured markdown plan with explicit task IDs and depends_on fields
- Use concise acceptance criteria and validation steps inside each task entry
- Include file locations and constraints to reduce ambiguity for subagents
- Run small task subsets for debugging before full-plan runs
- Ensure each subagent commits only its own changes and updates the plan immediately after completion
Example use cases
- Execute a repository onboarding plan that creates CI, linting, and basic tests in parallel waves
- Implement a multi-part feature where database migration must precede API changes and UI updates can run in parallel
- Run an incremental rollout plan: run backend tasks first, validate, then trigger frontend and docs tasks
- Process task subsets for hotfixes by specifying only critical task IDs and their dependencies
FAQ
The orchestrator inspects outputs, asks the subagent to retry or escalates the issue, and prevents dependent tasks from launching until the failure is resolved.
How are dependencies discovered?
Dependencies are parsed from each task's depends_on field in the markdown plan; if a subset is requested, the orchestrator includes required dependencies automatically.
Does the skill push commits?
No. Subagents are instructed to commit their changes locally and update the plan file; pushing is deliberately disallowed to avoid cross-agent conflicts.