skills.rest
alimtvnetwork
alimtvnetworkCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

execute-batched-loop-wor

Executes pending task queues via batched three-agent loops with chunked atomic git commits.

Updated Aug 10, 2026
Search Tags:#ci/cd#git workflow#task automation#multi-agent#batch execution#task queue#self-looping
One-click install
npx skills add https://github.com/alimtvnetwork/ui-prompts-cat --skill execute-batched-loop-wor-alimtvnetwork
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: execute-batched-loop-wor
Source: https://github.com/alimtvnetwork/ui-prompts-cat/tree/main/.agents/skills/execute-batched-loop-wor
Command: npx skills add https://github.com/alimtvnetwork/ui-prompts-cat --skill execute-batched-loop-wor-alimtvnetwork

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Coordinating large backlogs of pending development tasks often leads to merge conflicts, lost agent state on crashes, polluted git history, and wasted CI/CD credits from rapid polling. This Skill orchestrates a disciplined multi-agent execution loop that processes every pending task in .ai-memory/plans/pending/ without triggering releases or running tests locally. ## Core Features & Use Cases - Batched 3-Agent Execution: Spawns up to 3 concurrent sub-agents, each assigned small micro-task chunks with disjoint files tracked in a locking matrix to prevent merge collisions. - Crash Recovery & State Tracking: Logs agent state to .ai-memory/temp/ and .ai-memory/temp-agents/, applies a 3-strike rollback policy with git checkout, and records failures for diagnosis. - Strict Git Hygiene: Enforces single atomic commits (never per-file commits), artifact sanitization before staging, immediate push, and a total ban on local test/build execution deferred to CI/CD monitored via GitMap Pipeline-AI with ETA-based waiting. - Use Case: A repository accumulates 20 pending task files in .ai-memory/plans/pending/. Invoke this Skill to self-loop through all of them—grouping into execution waves, committing consolidated work, consolidating completed subtasks, and reporting done/pending statistics every loop—until the queue is empty. ## Quick Start Ask the AI to execute all pending tasks from the plans folder using the batched loop workflow without triggering a release, and say "continue" to resume the loop. ## Quick Start Run the batched loop to complete every pending task in .ai-memory/plans/pending/ without releases, then say go or continue to keep looping.

Frequently Asked Questions about execute-batched-loop-wor

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run pending tasks with multiple AI agents without merge conflicts?▼

Assign each sub-agent a chunk of micro-tasks touching completely disjoint files, and register active target files in a locking matrix such as active-locks.json. This Skill enforces a maximum of 3 concurrent agents with disjoint file sets to prevent git merge conflicts.

How do I resume the batched loop after it stops?▼

Say go, continue, or next to trigger another loop iteration. The Skill self-loops continuously until every task in .ai-memory/plans/pending/ is completed and moved to the completed folder.

Can I run tests or builds during routine task execution?▼

No. This workflow imposes a total ban on running tests (go test, pytest) and build checks (go build, npm run build) during routine execution. All verification is deferred to CI/CD, monitored via GitMap Pipeline-AI with ETA-based waiting.

What happens when a sub-agent crashes or fails repeatedly?▼

The agent's state is inspected in .ai-memory/temp/ or its isolated temp-agents folder. After 3 failures, dirty changes are reverted with git checkout, the root cause is logged to the plan and issues files, and a new agent restarts on a disjoint chunk.

Does this workflow trigger releases or version bumps?▼

No. This is a WOR (Without Release) workflow that strictly forbids version bumps, changelog edits, tags, and release scripts. Commits remain standard development commits pushed directly to the current branch.

Why must commits be atomic instead of per-file?▼

Per-file commits pollute git history, create subagent lock collisions, and break atomic rollback and bisectability. All modified files across a turn are accumulated and committed together in a single grouped commit, then pushed immediately.

Related Skills

Explore More Software Engineering Skills→
SeemSeamSeemSeam
17

su-batch

Authorize sequential batch execution of workflow subtasks with defined scope and stopping conditions.

Community
Advanced
thermiteauthermiteau
10

mav-bp-cicd-gitlab

Monitor GitLab CI/CD pipelines and retry failed jobs until they pass.

Official
Advanced
closedloop-aiclosedloop-ai
102

learning-quality

Capture and generalize actionable learnings from AI agent operations.

Official
Intermediate
marcusgollmarcusgoll

deep-execute

Processes bulk development task queues concurrently with Git conflict resolution and recovery support.

Community
Advanced