autonomous-coding-agent_skill

This skill helps you build autonomous coding agents and workflows by integrating Claude Code, multi-session management, and CI/CD pipelines.
  • TypeScript

14

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 shunsukehayashi/miyabi --skill autonomous-coding-agent

  • SKILL.md5.3 KB

Overview

This skill builds autonomous coding agents with CLI integration to drive programmatic code generation, multi-session workflows, and CI/CD automation. It exposes a safe execution pattern for calling Claude Code via a local CLI, session management primitives, and MCP tool definitions for integrating into larger pipelines. Use it to convert prompts into repeatable, auditable coding actions within TypeScript projects.

How this skill works

The skill wraps the Claude Code CLI in a server tool that spawns subprocesses, captures JSON output, and returns structured results. It includes safe execution patterns (timeouts, abort signals), prompt sanitization, working-directory validation, and explicit error handling. Multi-session classes let agents start, continue, and end sessions so stateful workflows can be orchestrated across runs or CI jobs.

When to use it

  • Automating code generation tasks in CI/CD pipelines or GitHub Actions
  • Running multi-step development workflows that require session continuity
  • Building autonomous agents that execute prompts programmatically
  • Creating audit-ready tooling that captures CLI output as JSON
  • Prototyping agentic workflows in TypeScript with MCP integration

Best practices

  • Sanitize and length-limit prompts before passing them to the CLI
  • Validate and restrict working directories to allowed sandboxes
  • Use timeouts and AbortController to avoid hung subprocesses
  • Return structured JSON from the CLI for reliable downstream parsing
  • Cover common error cases (timeout, CLI missing, parse errors) explicitly

Example use cases

  • A GitHub Action that fixes failing tests by running a Claude Code prompt and committing suggested changes
  • An MCP tool that executes code-mod transformations across multiple repositories in parallel
  • A session-based agent that iteratively refactors a large codebase across multiple prompts
  • A CI step that generates unit tests for changed code and outputs results as JSON
  • A local developer CLI that runs one-shot prompts and streams JSON results for tools integration

FAQ

Use a timeout parameter and AbortController support to cancel subprocesses; surface TIMEOUT errors and retry with adjusted prompts or resources.

How can I keep sessions stateful across CI steps?

Persist session identifiers returned by the execution result and pass them back into continue calls; implement session cleanup to avoid stale state.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational