When the agent drafts a spec in CrewPE, it also gives the draft a score out of ten. People ask two questions about that number: what does it measure, and can we just lock anything above nine? The second answer is no, and the reason explains the first.
The short answer
A readiness score is the agent’s own check of its draft spec: whether every requirement has testable acceptance criteria, whether scope is named, and which questions are still open. It tells a reviewer where to look. It does not mean the spec is right, and only a person can lock it.
Why the agent scores its own draft
Reviewing a spec is tiring. A reviewer reading a long draft tends to check what is there and miss what is not. The score exists to point at what is missing before a person spends time on the rest.
What it checks
Completeness
Does every requirement have at least one acceptance criterion? Are the repositories it touches named?
Testability
Can each criterion pass or fail? “Fast” and “user-friendly” lose points; numbers and observable results gain them.
Consistency
Do any two requirements contradict each other, or the locked intent brief they came from?
Open questions
What did the agent have to assume? Each assumption is listed as a question for the reviewer.
What it is not
It is not a measure of whether the spec is right. The agent can write a complete, testable, consistent spec for the wrong feature. It cannot know that the finance team actually needed a monthly report, not an export. Only the person who owns the problem can.
A high score means the spec is ready to review. It does not mean it is reviewed.
It is also not a target. If teams start writing to the score, they will produce specs that score well and say little. We keep it as a prompt for attention, not a pass mark.
How to use it well
- Read the open questions first. They are where the agent guessed, and where a wrong guess is cheapest to fix.
- Treat a low score as a conversation. It usually means the intent was unclear, not that the agent wrote badly.
- Never lock on the number alone. Read the requirements as the person who will accept the result.
Why a person still locks
Locking a spec is a decision: this is what we will build, and I am accountable for it. A score cannot take that accountability, so it never locks anything. It only makes the person who does a little faster and a little harder to fool.
In CrewPE
The Thread Analyst drafts the spec from the locked intent brief, scores its own draft and lists its open questions. You decide when it is locked.
See how CrewPE worksQuestions teams ask
- What is a spec readiness score?
- A score the drafting agent gives its own spec, based on whether requirements have testable acceptance criteria, whether scope and repositories are named, whether anything contradicts, and which questions remain open.
- Can a spec be approved automatically if the score is high?
- No. A high score means the draft is ready for a person to review. Only the person who owns the outcome can judge whether it describes the right thing, so only a person locks it.
- What does a low readiness score usually mean?
- Usually that the original request was unclear. The open questions show what the agent had to guess, and answering them often raises the score more than rewriting the draft.
Parani Raja
Co-founder of EnWithAI, working on legacy modernization: keeping what old systems know when they are rebuilt.