to-PRD: Turn Finished Work into Review-Ready Specs

Where this fits

You just merged a PR with context scattered across commits, or you're staring at a Linear ticket that says "improve onboarding" and nothing else. Relying on your agent to regurgitate the diff produces noise, not a document product stakeholders will read. This skill steps in: it takes finished work (or a planned feature) and turns it into a structured PRD using the CopilotKit EEP spec template—without touching RFCs, implementation plans, or your Linear board.

Why agents benefit

• Grills before filling: The skill reads assets/prd-template.md in full and asks clarifying questions instead of blindly populating sections with invented scope.
• Finds facts itself: The agent is instructed to dig through tickets, PRs, and commit history to gather context rather than pestering you for details it can retrieve.
• Writes complete user stories: Covers every actor and path—empty states, error cases, admin workflows—so reviewers rarely bounce a draft back for missing edge cases.
• Keeps implementation readable: Names modules, interfaces, schema, and API contracts without dumping file paths or code blocks, except a short prototype snippet where prose fails.
• No accidental side-effects: It writes the spec literally. No git push, no implementation scaffolding, no Linear status changes.

Practical scenarios

Post-merge cleanup

Your team shipped a feature across three stacked PRs. [agent] /to-prd reads the diff, pulls commit messages, and produces a PRD that reflects what actually landed—not what was originally planned.

Triage for a vague ticket

A PM drops a one-liner: "improve onboarding." The skill parses related Linear issues and past PRs, assembles the user stories, then asks you only the few gaps it genuinely could not infer.

Feature handoff before kickoff

You have a feature description sitting in a Slack thread. You invoke the skill, and the agent emits a structured spec with implementation decisions, ready for team review without you rewriting a bare outline.

Add it to your agent workflow

Add the skill to your runtime:

npx skills add AlemTuzlak/skills --skill to-prd

Then trigger it in chat:

/to-prd — Draft a PRD from the merged branch feature/checkout-flow

The skill reads the template, asks a couple of clarifying questions if needed, and outputs the full EEP-format spec. To see how it fits among similar agent-building tools, browse the directory at https://www.openagentskill.com/skills/alemtuzlak-to-prd.

Compare before adopting

Before making it a default, check the quality signals: does the EEP template match your team's spec shape, or will you trim output every time? Look at maintenance freshness—last commit date and issue responses matter for a skill that depends on a template file. Compare against the simplest alternative: your own prompt plus a carefully written template may suffice for occasional drafts. Finally, confirm workflow fit: the "grill first" and "never touch Linear" constraints are deliberate, and they may conflict with agents that prefer to auto-categorize or auto-close.

Why it is worth tracking

Track this if your team runs on spec-heavy feature work. The skill is mostly a template plus inference rules, so maintenance cost is low, and the 39-star count signals steady interest rather than viral noise. Re-evaluate it whenever your product template changes or you introduce Linear automation that could collide with its careful no-side-effects boundary.