Does PRD help? Tested on spec writing tasks

PRD, from github/awesome-copilot. Its spec writing tasks were run with the skill loaded and without it. Each time it is the same task, run twice.

What we tested

Whether loading this skill helps a model write a product spec.

What counts as helping

The score with the skill has to beat the score without it by at least 10 percentage points, on the same 10 briefs. A blind reader has to prefer the version written with the skill at least 60 times in 100, across the same 10 briefs.

How we scored it

We checked each spec against a fixed list of things a spec needs, and we also had a blind reader compare the two versions. The two are reported separately and never added together.

All 6 models tested already passed this without the change, and still passed with it.

There is no chart here because there is no shape to draw: every value sits at one end of the scale, the models agree, and no range of likely values is wide enough to see. The per-model numbers are in the tables below.

Show per-task detail

Every item, without and with, per model

Via API tested via API

  • GPT-5 mini Deterministic
  • Gemini 3.1 Flash Lite Deterministic

deterministic pass rate, 0 to 1

Hover or focus an item to read its task and every model’s two values.

Task-paired: these item means are what this cell's interval is built from.

The item means behind this plot
Per-item arm means. Without is the unaided arm, with is the treated arm.
ItemGPT-5 mini Deterministic Gemini 3.1 Flash Lite Deterministic
Task without with without with
sw-01 0.909 0.909 1.000 1.000
sw-03 0.909 0.182 1.000 1.000
sw-05 0.909 0.909 1.000 1.000
sw-07 0.909 1.000 1.000 1.000
sw-08 0.909 1.000 1.000 1.000
sw-02 1.000 1.000 1.000 1.000
sw-04 1.000 1.000 1.000 1.000
sw-06 1.000 1.000 1.000 1.000
sw-09 1.000 1.000 1.000 1.000
sw-10 1.000 0.909 1.000 1.000
A picture of the per-item means behind the per-model numbers. The tables are the source. Each model draws two lines over the same task set: a dashed line through its unaided scores and a solid line through its treated ones. Items are ordered by the mean unaided score across the models that measured them, lowest first. The two instruments are reported separately and never averaged.

Without to with, per model

In Claude Code tested in Claude Code

Claude Opus 5 Expansion cohort (three arms)

No measured effect

Claude Fable 5.1 Expansion cohort (three arms)

No measured effect

Claude Sonnet 5 Expansion cohort (three arms)

No measured effect

Claude Haiku 4.5 Expansion cohort (three arms)

No measured effect

deterministic pass rate, 0 to 1

A picture of the per-model numbers, drawn from the same results. The tables are the source. Each row runs from the score without to the score with. The thin bar beneath it is the bracket, and it shows how much the difference could move if we ran it again. The two test methods are reported separately and never averaged. A bracket wider than the axis is drawn to the edge with its end cap omitted; the table gives its bounds.

The Claude Code runs committed their cells as arm means and intervals rather than as per-item values, so this panel is drawn without to with rather than item by item.

What exactly was tested, and how it was scored
Repository
github/awesome-copilot
Path
skills/prd
Commit
c956566a35c3c2e635f019e7a1bfa59d9497e8b1
Content hash
fc1caf7b8c27e187d50b76ae52b0972bbf95b7bbc16e39a4d138a132bd76aea8
Date tested
not yet tested

The pin is the whole of this skill's identity here. It resolves at https://github.com/github/awesome-copilot/tree/c956566a35c3c2e635f019e7a1bfa59d9497e8b1/skills/prd, and the content hash is a sha256 over exactly the text the model was given with the skill loaded. Nothing else about the skill appears on this site.

S19-github-prd

Loading skills/prd from github/awesome-copilot improves outputs on spec writing tasks.

Pass criterion
With-arm mean structure score exceeds the without-arm by at least 0.10, and the with-arm win rate is at or above 0.60, on the same 10 briefs.
Scale
unit. Same scorer, same instrument and same scale as S06-product-capability. Added 2026-08-31 by the source expansion under R23; adding a row to an instrument does not touch the instrument.
Task pairs planned per model
10
Notes
Shares the spec-writing task set with every other claim in this class, so all of its rows are measured on identical items against one answer key. The slot this claim scores was pinned by the 2026-08-31 source expansion under R23 and NO MODEL HAS BEEN RUN AGAINST IT. The claim exists so the class holds one claim per pinned slot, which is what makes the row comparable the day it is run.

Injected context tokens

Injected context tokens, per arm
ArmContext charactersInjected context tokens
without00
with4,3281,016

The character count is exact: it is the length of the text the with arm is given, and the content hash above is a sha256 over that same text. The token figure is an estimate at 4.26 characters per token, the ratio the phase 1 run measured over 3,120 calls, and it is labelled an estimate until a run reports its own token counts. The without arm is given the identical prompt and nothing else, so its zero is a measurement rather than a missing value.

Verdict per model

Effect per model
ModelReadingOrbitWithout -> withEffect95 percent intervalTask pairsModel version returned
Gemini 3.1 Flash LiteDeterministic, scored against a committed answer keyUnobservable (b) both arms at a bound, the task set did not separate them1.0000 -> 1.0000+0.0000[0.0000, 0.0000]10 pairs, 30 of 30 recordsgemini-3.1-flash-lite
GPT-5 miniDeterministic, scored against a committed answer keyIn free drift (10 of 10 pairs) Nothing left to measure. The model scored 95% without the skill.0.9545 -> 0.8909-0.0636[-0.2116, 0.0844]10 pairs, 30 of 32 recordsgpt-5-mini-2025-08-07, gpt-5-mini
  • Unobservable: the cell cannot be classified: an unbounded metric, or both arms at a bound of the scale.
  • In free drift: no separation the design can resolve. Not evidence of no effect.

Cost

Defined in the metrics canon. Cost per task divides every dollar spent on an arm by the tasks attempted on it, including tasks whose call returned nothing, because a call that returned nothing was still billed.

Where the with arm did not win

Gemini 3.1 Flash Lite, Deterministic, scored against a committed answer key

Drew on 10 of 10 pairs: sw-01, sw-02, sw-03, sw-04, sw-05, sw-06, sw-07, sw-08, sw-09, sw-10.

GPT-5 mini, Deterministic, scored against a committed answer key

Lost on 2 of 10 pairs: sw-03 (-0.7273), sw-10 (-0.0909).

Drew on 6 of 10 pairs: sw-01, sw-02, sw-04, sw-05, sw-06, sw-09.

In Claude Code

These cells are tested in Claude Code, on a subscription path with no API key. They are a second instrument: no figure here is averaged with one tested via API above. How the two were compared.

  • Claude Fable 5.1 · unclear in the three-arm cohort
  • Claude Haiku 4.5 · unclear in the three-arm cohort
  • Claude Opus 5 · unclear in the three-arm cohort
  • Claude Sonnet 5 · unclear in the three-arm cohort

Expansion cohort (three arms)

one pass over each committed item. Matrix hash 0078cfa8. Pre-registration · Run report.

One row per model. Every figure is read from this run’s committed cells.
ModelNo skillWith skillDeltaIntervalPairsOrbitInjected tokens
Claude Fable 5.10.95450.9455-0.009-0.051 to 0.03210In free driftnot measured: this run records the dose per arm, in the doses list
Claude Haiku 4.50.97270.9909+0.018-0.006 to 0.04210In free driftnot measured: this run records the dose per arm, in the doses list
Claude Opus 50.90910.9091+0.0000.000 to 0.00010In free driftnot measured: this run records the dose per arm, in the doses list
Claude Sonnet 50.95450.9727+0.018-0.026 to 0.06310In free driftnot measured: this run records the dose per arm, in the doses list

Every figure above is computed at build time from harness/results/runs-skills.jsonl and harness/results/ledger-skills.jsonl, both committed, by the frozen status_v1 rule and the metrics_v1 cost definitions. How a claim gets tested.

720 run records behind this page. Every verdict is computed at build time by the same frozen status_v1 rule that decides every other verdict on this site, and nothing here is written by hand.