Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Turns the current conversation into a spec and publishes it to the project issue tracker, with no interview. Use when the user says "write a spec", "create a PRD", or wants the discussion captured as a spec.
.claude/skills/fradser-to-spec/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-06 | ✗→✓ | ▲ Improved | 78% | 0% |
| case-08 | ✓→✗ | ▼ Worse | -73% | 0% |
| case-12 | ✓→✗ | ▼ Worse | -65% | 0% |
| case-15 | ✓→✗ | ▼ Worse | -69% | 0% |
| case-17 | ✓→✗ | ▼ Worse | -69% | 0% |
This skill takes the current conversation context and codebase understanding and produces a spec (you may know this document as a PRD). Do NOT interview the user — just synthesize what you already know.
The issue tracker and triage label vocabulary should have been provided to you — run /mattpocock:setup-matt-pocock-skills if not.
Do NOT interview the user; the spec is a synthesis of the conversation and codebase understanding already in context. Test at the highest existing seam, preferring one seam to many — check the seams with the user before writing.
Check with the user that these seams match their expectations.
ready-for-agent triage label - no need for additional triage.<spec-template>
The problem that the user is facing, from the user's perspective.
The solution to the problem, from the user's perspective.
A LONG, numbered list of user stories. Each user story should be in the format of:
<user-story-example>
</user-story-example>
This list of user stories should be extremely extensive and cover all aspects of the feature.
Executable scenarios derived from the user stories above, written as Gherkin (Given/When/Then). Each scenario pins one behavior; collectively they are the executable specification the implementation is verified against. See /mattpocock:bdd for the scenario authoring guide and the .feature storage convention.
A list of implementation decisions that were made. This can include:
Do NOT include specific file paths or code snippets. They may end up being outdated very quickly.
Exception: if a prototype produced a snippet that encodes a decision more precisely than prose can (state machine, reducer, schema, type shape), inline it within the relevant decision and note briefly that it came from a prototype. Trim to the decision-rich parts — not a working demo, just the important bits.
A list of testing decisions that were made. Include:
A description of the things that are out of scope for this spec.
Any further notes about the feature.
</spec-template>
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 17,485 | 15,295 | -13% | 1 | 1 | 0% | 2,624 | 1,034 | -61% | 0 | 0 | — |
case-02 | fail→fail | 12,683 | 3,856 | -70% | 1 | 1 | 0% | 2,061 | 1,004 | -51% | 0 | 0 | — |
case-03 | fail→fail | 7,842 | 4,678 | -40% | 1 | 1 | 0% | 1,172 | 1,097 | -6% | 0 | 0 | — |
case-04 | fail→fail | 19,025 | 5,376 | -72% | 1 | 1 | 0% | 2,695 | 1,108 | -59% | 0 | 0 | — |
case-05 | fail→fail | 19,504 | 5,431 | -72% | 1 | 1 | 0% | 3,108 | 1,000 | -68% | 0 | 0 | — |
case-06 | fail→pass | 19,428 | 28,332 | +46% | 1 | 1 | 0% | 3,196 | 5,691 | +78% | 0 | 0 | — |
case-07 | fail→fail | 17,271 | 6,108 | -65% | 1 | 1 | 0% | 3,723 | 1,145 | -69% | 0 | 0 | — |
case-08 | pass→fail | 20,834 | 3,673 | -82% | 1 | 1 | 0% | 3,480 | 934 | -73% | 0 | 0 | — |
case-09 | fail→fail | 21,115 | 4,520 | -79% | 1 | 1 | 0% | 3,324 | 1,006 | -70% | 0 | 0 | — |
case-10 | fail→fail | 10,062 | 4,793 | -52% | 1 | 1 | 0% | 1,506 | 970 | -36% | 0 | 0 | — |
case-11 | fail→fail | 16,042 | 4,689 | -71% | 1 | 1 | 0% | 2,525 | 988 | -61% | 0 | 0 | — |
case-12 | pass→fail | 17,870 | 4,213 | -76% | 1 | 1 | 0% | 2,911 | 1,009 | -65% | 0 | 0 | — |
case-13 | fail→fail | 19,086 | 4,132 | -78% | 1 | 1 | 0% | 3,234 | 1,009 | -69% | 0 | 0 | — |
case-14 | fail→fail | 17,648 | 6,012 | -66% | 1 | 1 | 0% | 3,240 | 1,087 | -66% | 0 | 0 | — |
case-15 | pass→fail | 20,154 | 5,102 | -75% | 1 | 1 | 0% | 2,991 | 933 | -69% | 0 | 0 | — |
case-16 | fail→fail | 15,403 | 3,943 | -74% | 1 | 1 | 0% | 2,793 | 1,114 | -60% | 0 | 0 | — |
case-17 | pass→fail | 18,566 | 4,855 | -74% | 1 | 1 | 0% | 3,507 | 1,075 | -69% | 0 | 0 | — |
case-18 | pass→fail | 18,436 | 4,478 | -76% | 1 | 1 | 0% | 3,347 | 1,017 | -70% | 0 | 0 | — |
case-19 | pass→fail | 3,125 | 5,486 | +76% | 1 | 1 | 0% | 481 | 1,242 | +158% | 0 | 0 | — |
case-20 | fail→fail | 3,365 | 5,919 | +76% | 1 | 1 | 0% | 520 | 1,085 | +109% | 0 | 0 | — |
case-21 | pass→fail | 13,639 | 3,033 | -78% | 1 | 1 | 0% | 3,342 | 915 | -73% | 0 | 0 | — |
case-22 | pass→fail | 7,710 | 4,457 | -42% | 1 | 1 | 0% | 1,061 | 1,126 | +6% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 1 counted toward the lift figure. The other 21 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of -32 percentage points is the difference between those two pass rates over the 1 comparable cases. 11 cases got worse with the skill loaded, and they are included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.