okhp3-replit-task-executor

Execute Replit-planned tasks with an external coding agent through protected Git integration.

Updated Aug 17, 2026
One-click install
npx skills add https://github.com/OKHP3/telling-forward --skill okhp3-replit-task-executor-okhp3
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: okhp3-replit-task-executor
Source: https://github.com/OKHP3/telling-forward/tree/main/.agents/skills/okhp3-replit-task-executor
Command: npx skills add https://github.com/OKHP3/telling-forward --skill okhp3-replit-task-executor-okhp3

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) and assets (resource) components.

What problem does it solve? Coordinating work between Replit's planning agent and an external coding agent risks duplicate implementation, lost unpublished commits, false task completion, and unauthorized spending. This Skill defines a responsibility contract and guarded workflow so Replit stays project manager and designer while an external agent owns architecture, code, tests, and Git integration. ## Core Features & Use Cases - Bounded task handoff: Capture one selected task or a frozen backlog batch into a reusable task packet with original acceptance criteria, dependencies, and authorized outcomes. - Ownership and race protection: Establish a single writer before editing, bind merges to a reviewed head SHA with expected-head guards, and block integration when ownership or base state is unresolved. - Verified completion return: Map every criterion to evidence in a completion receipt, keep design acceptance, board disposition, CI, and deployment as independently reported statuses, and never fabricate board-completion APIs. - Use Case: A product owner has a Replit-planned backlog and wants Claude Code or Codex to implement one approved task. The external agent captures the task packet, verifies no other writer owns the files, implements on a feature branch, opens a protected PR, and returns a receipt for bounded Replit design acceptance without triggering another paid build. ## Quick Start Ask the agent to execute the selected Replit task externally using the okhp3-replit-task-executor skill and return a verified completion receipt.

Frequently Asked Questions about okhp3-replit-task-executor

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

FAQPage Schema
How do I hand off a Replit task to an external coding agent?▼

Capture the selected task into the task packet with its full acceptance criteria, dependencies, and authorized outcome. Then establish single-writer ownership, implement on a feature branch, run project gates, and open a protected PR before returning a completion receipt.

How do I avoid duplicate work between Replit Agent and an external executor?▼

Inspect the task state, nearby tasks, open PRs, and worktrees before editing, and record a shared ownership claim. Queued, Active, or quota-blocked tasks remain owned by Replit, so overlapping edits return ownership-blocked until the owner resolves the writers.

Can I mark a Replit task complete from the Shell or a Git commit?▼

No. Root replit.md, .replit configuration, Git task trailers, and historical markTaskComplete calls are provenance, not a callable completion API. Prepare a completion receipt and use only a currently supported, authorized administrative route, otherwise leave board-update-pending.

Does this skill work with ChatGPT or Claude chat without coding tools?▼

Chat-only clients can prepare a complete task packet for a coding-capable host, but execution is reported not-run. They must not invoke Replit Agent prompt tools like update_app_using_prompt as a substitute implementation route.

What happens when a PR head or target base changes before merge?▼

A changed head or relevant base movement invalidates the prior review evidence. The workflow requires an expected-head SHA guard or verified serialized handoff; if no supported protection exists, integration is deferred rather than risking a stale merge.

What are the limitations of this Replit executor workflow?▼

Maturity remains draftable: there is no measured cost improvement, no completed current-version live end-to-end pilot, and no cross-host production benchmark. It also cannot cancel cards, buy usage, or force board completion by default.