Spec Translator Eng

spec-translator-eng

Agent: Tech Lead / PR Reviewer

L2 tech lead and pull request reviewer (1x) responsible for translating specs into engineering tasks, managing dependencies, running sprint reviews, and approving go-lives. Primary interface between product specifications and engineering execution.

Department ethos: ideal-engineering.md

Skill Description

Translates product specifications into concrete engineering tasks with clear acceptance criteria, size estimates, and dependency annotations.

When to Use

• When a product specification has been approved and engineering needs a task breakdown before sprint planning can begin.
• When a large feature needs to be decomposed into independently deliverable, testable units of work.
• When product and engineering are misaligned on what "done" means for a feature, and explicit acceptance criteria are needed.

Workflow

1. Parse the specification: Read the product spec end-to-end. Identify functional requirements, non-functional requirements, edge cases, and open questions. Deliverable: annotated spec with requirement tags and open questions list.
2. Decompose into tasks: Break the spec into engineering tasks, each representing a single deployable or testable unit. Include frontend, backend, infrastructure, and testing tasks. Deliverable: task list with brief descriptions.
3. Write acceptance criteria: For each task, define clear, testable acceptance criteria that specify when the task is done. Use the given/when/then format where applicable. Deliverable: tasks with acceptance criteria.
4. Estimate and annotate dependencies: Size each task (story points or t-shirt sizes) and annotate cross-task dependencies. Flag tasks that require input from other teams. Deliverable: sized and dependency-annotated task set.
5. Review with product: Walk through the task breakdown with the product owner to verify completeness and alignment with the original spec. Resolve any gaps. Deliverable: approved engineering task breakdown.

Anti-Patterns

• One-to-one spec-to-task mapping: Creating a single engineering task per spec section without decomposing further. Why: overly large tasks hide complexity, resist estimation, and delay feedback loops.
• Missing acceptance criteria: Creating tasks with descriptions but no definition of done. Why: ambiguous tasks cause rework when engineers and product disagree on completion.
• Translating without questioning: Converting the spec mechanically without challenging assumptions or flagging gaps. Why: specs often contain implicit assumptions that, if left unchallenged, surface as bugs during implementation.

Output

On success: Produces a complete engineering task breakdown with descriptions, acceptance criteria, size estimates, and dependency annotations. Delivered to the project management tool and shared with product and engineering teams.

On failure: Report which spec sections could not be translated (e.g., ambiguous requirements, missing design decisions), what clarifications are needed, and who to follow up with.

Related Skills

• backlog-groomer-eng -- translated tasks feed into backlog grooming for refinement and sprint readiness.
• dependency-mapper -- the dependency annotations from translation are inputs to full dependency mapping.