Write a feature specification with AI
Turn decisions already made in a project conversation into a feature specification. Review the testing boundaries and publish the agreed scope to the configured tracker.
To write a feature specification with AI, start from a conversation in which the major decisions are already settled. The to-spec skill reads the relevant project context, proposes testing boundaries and writes the agreed feature into the configured tracker.
For a cancellation feature, the important details may be who can cancel, when the option disappears and what the buyer sees afterwards. The specification can capture those behaviours and identify what is outside the current release without copying transient implementation details into every story.
What it does
- Describe the problem and solution from the user’s perspective.
- Capture agreed user stories and implementation decisions.
- Identify a suitable boundary for behavioural tests.
- Record deliberately excluded work.
- Publish the specification with the project’s agreed triage label.
Install and use
- Install the skill and provide repository access and the configured tracker. Use the maker’s setup companion if the destination and label vocabulary are missing.
- Ask “Turn our agreed cancellation behaviour into a feature spec. Propose the highest useful boundary for testing it.”
- Review the proposed testing scope, then inspect the published specification and its triage label in the intended project.
FAQ
Source
Made by Matt Pocock. Original skill: to-spec, published on GitHub under the MIT licence.
View on GitHub