A task file stored in Git lets AI agents read requirements and code in one place.
TASKS AS CODE REQUIRES MOVING JIRA TICKETS INTO GIT REPOS
*AI coding agents need the same repository for code, docs, and tasks. Storing tickets as Markdown in Git promises versioned, reviewable work and tighter control. The shift challenges entrenched Jira workflows.*
The software world is waking up to a simple truth: AI coding agents live inside the repository. Yet most teams still silo tickets in Jira or Redmine, forcing agents to chase a moving target. The article "Tasks as Code: Move Your Tickets into the Repository" argues that storing tasks as Markdown files in Git aligns requirements, code, and documentation under one roof. This alignment promises versioned, reviewable work and a single source of truth for both humans and machines. The shift mirrors the Docs as Code movement, which already treats documentation as code. By extending the model to tickets, organizations can let AI agents read, plan, and report on work without leaving the repository.
Why move tasks into Git?
AI agents read the repository as their source of truth. When tasks live as Markdown files next to code and documentation, the agent can ingest requirements, examine the relevant code, and produce changes without hopping between tools. Every task change becomes a commit, inheriting Git's versioning and review mechanisms. The Docs as Code movement already places documentation in a /docs folder inside the repo, a practice known as Docs Like Code. Extending that model to tickets eliminates the disconnect between issue trackers like Jira and the code base, turning tasks into first‑class artifacts that are searchable, branchable, and auditable.
How does the five‑step workflow operate?
The workflow begins with a short task description in Markdown. An AI agent then reads the file, analyzes the surrounding code and docs, and drafts an implementation plan that lists proposed changes. A developer reviews that plan and explicitly authorizes it. Only after approval does the agent modify the code, update any affected documentation, and append an implementation report to the same task file. The report, plan, and original description stay together, making it trivial to trace requirements to code changes. The approval gate ensures humans retain control while the agent handles the heavy lifting.
What security benefits come from separate repositories?
Documentation often needs to be public, but task files may contain unfinished security fixes, internal roadmaps, or unreleased features. By storing code, docs, and tasks in three distinct Git repositories, teams can assign independent access controls. The task repository can remain private while the documentation repo stays open. Locally, developers clone the docs and tasks into folders inside the code repo and list those folders in .gitignore, so the parent repo does not track their contents. Each repository preserves its own history, allowing secure, granular synchronization without exposing sensitive tickets.
Can existing tools like Typemill support this shift?
Typemill synchronizes content with Git and handles multiple repositories. It can host a public documentation project alongside a private task project, presenting both through a single browser interface while keeping the underlying files in separate repos. The interface lets users browse, edit, and organize Markdown files without replacing Git as the version control system. Because the workflow only requires Git and an AI coding agent, adding Typemill provides a convenient UI layer without altering the core process.
Frequently asked questions
What is Tasks as Code?
Tasks as Code stores tickets as Markdown files in a Git repository, placing them next to code and documentation so AI agents can read and act on them.
How does the five‑step workflow keep humans in control?
The workflow includes a review and approve step where a developer checks the AI‑generated implementation plan before the agent makes any code changes.
Why keep tasks in a separate Git repository?
Separate repositories allow independent access controls, keeping public documentation open while protecting private task files that may contain security fixes or unreleased features.
Adopting Tasks as Code forces a hard look at legacy issue trackers. The approach leverages Git's proven audit trail, gives developers a clear approval checkpoint, and isolates sensitive work in private repos. With tools like Typemill providing a thin UI layer, the transition requires no new infrastructure—just a disciplined move of Markdown files. Teams that act now will let AI agents operate at full speed while keeping humans firmly in the loop.
Sources: [Hacker News – Tasks as Code: Move Your Tickets into the Repository]