About The Agentic Sprint

The Agentic Sprint is a proposed operating model for software delivery in which autonomous agents execute bounded engineering work and people keep authority over intent, architecture, acceptance and release. This site is its home.

What this site is

The manifesto is the front page: ten principles and a compact test. Behind it sit 13 numbered documents, D1 to D13, and 9 working templates, T1 to T9. The specification (D1) is the normative core: it states what a conforming Agentic Sprint requires, using the RFC 2119 keywords. The companion standards scope that core; the informative documents explain, measure and apply it without adding requirements; the templates are the artefacts a team fills in.

Every document carries a stable identifier, a status, a version and the documents it builds on. Cite a requirement by its document and identifier, not by a page position.

Authorship

The methodology was written by Siri Dalugoda, a technology executive and researcher based in Auckland, New Zealand. It is published independently. No organisation owns it, and nothing here should be read as a claim about the practices of any employer or client.

Status

The series is at version 0.1. It is a proposed methodology, offered for adoption, pilots and critique. It is not a claim of industry consensus, a safety certification, or evidence that current AI systems can reliably perform all professional engineering work without supervision. Changes are versioned and dated on each document, and a new version of a document never silently replaces the requirements of the one before it.

Licence

• The documents, D1 to D13, are published under CC BY 4.0. Share and adapt them for any purpose, including commercially, with attribution.
• The templates, T1 to T9, are dedicated to the public domain under CC0 1.0. A team may copy them into a repository without attribution.
• The source code of this site is licensed under Apache 2.0 and is on GitHub.

None of these licences grants the right to imply that a derived work is endorsed by, or is the official work of, the author.

How to cite

To cite the methodology as a whole:

To cite a single document, use the citation at the foot of that document. It names the document, its identifier, its version and its date, so a reader can tell which text was being referred to.

How to contribute

The site and its content are maintained in the open. Open an issue or a pull request on GitHub to propose a correction, argue with a requirement or report an implementation experience. Proposals that change a normative requirement are handled as a new version of the affected document, with the change dated.

Relationship to dalugoda.com

The methodology was first published on the author’s site, dalugoda.com. This site is now the canonical home of every document and template. Two related pieces remain on dalugoda.com and are not part of the methodology proper: the original essay, which introduced the model in plain language, and Autonomous Loop, an open-source implementation of the optional inner execution loop that D4 describes.

For machines

Every document has a clean Markdown edition at /md/<slug>, linked from the page as its text/markdown alternate. The site is indexed for retrieval agents at /llms.txt and /llms-full.txt, and publishes a sitemap, RSS and Atom feeds. Each page carries structured data naming the document, its identifier, its version and its place in the series.