Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Amazon Working Backwards PR/FAQ generator that forces customer-outcome thinking before any code is written. Produces a press release, an internal FAQ, and an external FAQ as a single decision artifact.
.claude/skills/borghei-prfaq/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-04 | ✗→✓ | ▲ Improved | 103% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 122% | 0% |
| case-09 | ✗→✓ | ▲ Improved | 40% | 0% |
| case-10 | ✗→✓ | ▲ Improved | 228% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 64% | 0% |
The PR/FAQ is Amazon's "working backwards" artifact: before any team is funded to build a product, the PM writes a future-dated press release and an FAQ that anticipates every hard question. The discipline forces clarity on customer, problem, and outcome before a single design decision is made. If the team cannot write a compelling, credible PR/FAQ, the idea is not ready.
Use this skill when you need a high-fidelity narrative artifact for funding review, executive sponsorship, or a "should we even do this?" decision. The output is a single Markdown document with three sections: the press release, an internal FAQ (the questions your CFO, lawyer, and head of engineering will ask), and an external FAQ (the questions customers and press will ask). The PR/FAQ is not a PRD substitute -- it precedes the PRD -- and is not a marketing draft. The "Press Release Test" is the bar: if the press release would not strike a customer as genuinely newsworthy, the idea needs more work.
When NOT to use: incremental features on an existing product (use wwas/ or job-stories/); a signed-off PRD mid-build; pure technical infrastructure with no end-customer narrative.
Before generating the PR/FAQ, confirm these inputs. If any is unknown or vague, ASK — do not assume:
Stop rule: ask only the 2-3 that most change the output. If the user says "just draft it," proceed and list your assumptions at the top of the artifact.
create-prd/.Use assets/prfaq_template.md as the fill-in scaffold. See references/prfaq-playbook.md for the full structure, FAQ categories, sample questions, workflow, and troubleshooting.
assets/prfaq_template.md — complete fill-in PR/FAQ template (PR + internal FAQ + external FAQ).In scope: press-release drafting in the Working Backwards format; internal FAQ across the 9 categories; external buyer/customer FAQ; the Press Release Test protocol; handoff to PRD (create-prd/) and roadmap (outcome-roadmap/).
Out of scope: marketing/ad copy or external launch press (handoff to marketing); detailed PRD authoring (create-prd/); OKR definition (brainstorm-okrs/); backlog drafting (wwas/, job-stories/); pricing strategy or financial modeling (finance/).
Caveats: The PR/FAQ is an internal alignment artifact, not for external publication. The format works best in cultures that tolerate written narrative memos; slide-deck cultures may need to adopt it gradually. A weak PR/FAQ is a feature, not a bug — it surfaces a weak idea; fix the idea, do not inflate the language.
| Integration | Direction | Description | |-------------|-----------|-------------| | discovery/brainstorm-ideas/ | Receives from | Top-ranked opportunity solutions become PR/FAQ candidates | | discovery/identify-assumptions/ | Receives from | Assumptions populate internal FAQ "what we do not know yet" answers | | discovery/pre-mortem/ | Receives from | Tiger risks populate internal FAQ "failure modes" answers | | execution/create-prd/ | Feeds into | Approved PR/FAQ becomes the prologue and scope anchor for the PRD | | execution/brainstorm-okrs/ | Feeds into | Press release outcomes become OKR candidates | | execution/outcome-roadmap/ | Feeds into | PR/FAQ launch date and v1 scope inform roadmap horizon placement | | execution/roadmap-communication/ | Feeds into | PR/FAQ summary becomes the executive-variant roadmap narrative | | marketing/ (skills domain) | Hands off to | After launch approval, marketing teams adapt the PR into actual launch press materials |
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-04 | fail→pass | 18,987 | 26,653 | +40% | 1 | 1 | 0% | 2,675 | 5,429 | +103% | 0 | 0 | — |
case-05 | pass→pass | 22,987 | 24,643 | +7% | 1 | 1 | 0% | 3,369 | 4,915 | +46% | 0 | 0 | — |
case-14 | fail→pass | 18,149 | 32,615 | +80% | 1 | 1 | 0% | 2,748 | 6,096 | +122% | 0 | 0 | — |
case-01 | fail→fail | 29,584 | 27,919 | -6% | 1 | 1 | 0% | 4,163 | 5,528 | +33% | 0 | 0 | — |
case-02 | fail→fail | 24,833 | 31,454 | +27% | 1 | 1 | 0% | 3,561 | 6,027 | +69% | 0 | 0 | — |
case-03 | fail→fail | 33,387 | 39,734 | +19% | 1 | 1 | 0% | 4,910 | 7,199 | +47% | 0 | 0 | — |
case-06 | pass→pass | 18,162 | 23,440 | +29% | 1 | 1 | 0% | 2,875 | 4,511 | +57% | 0 | 0 | — |
case-07 | pass→pass | 11,887 | 21,118 | +78% | 1 | 1 | 0% | 1,897 | 4,534 | +139% | 0 | 0 | — |
case-08 | pass→fail | 24,241 | 10,274 | -58% | 1 | 1 | 0% | 4,547 | 3,109 | -32% | 0 | 0 | — |
case-09 | fail→pass | 13,553 | 8,739 | -36% | 1 | 1 | 0% | 1,990 | 2,786 | +40% | 0 | 0 | — |
case-10 | fail→pass | 5,148 | 5,504 | +7% | 1 | 1 | 0% | 710 | 2,328 | +228% | 0 | 0 | — |
case-11 | fail→pass | 11,837 | 11,838 | +0% | 1 | 1 | 0% | 1,939 | 3,177 | +64% | 0 | 0 | — |
case-12 | pass→pass | 13,279 | 14,464 | +9% | 1 | 1 | 0% | 2,103 | 3,588 | +71% | 0 | 0 | — |
case-13 | pass→pass | 13,841 | 14,499 | +5% | 1 | 1 | 0% | 1,901 | 3,391 | +78% | 0 | 0 | — |
case-15 | pass→pass | 10,570 | 5,020 | -53% | 1 | 1 | 0% | 1,599 | 2,307 | +44% | 0 | 0 | — |
case-16 | fail→pass | 24,765 | 7,343 | -70% | 1 | 1 | 0% | 1,111 | 2,601 | +134% | 0 | 0 | — |
case-17 | pass→pass | 9,219 | 7,122 | -23% | 1 | 1 | 0% | 1,304 | 2,537 | +95% | 0 | 0 | — |
case-18 | pass→pass | 8,196 | 6,870 | -16% | 1 | 1 | 0% | 1,172 | 2,466 | +110% | 0 | 0 | — |
case-19 | pass→pass | 18,232 | 9,957 | -45% | 1 | 1 | 0% | 2,569 | 3,064 | +19% | 0 | 0 | — |
case-20 | pass→pass | 12,768 | 9,913 | -22% | 1 | 1 | 0% | 1,901 | 2,896 | +52% | 0 | 0 | — |
case-21 | fail→pass | 11,762 | 5,608 | -52% | 1 | 1 | 0% | 1,696 | 2,295 | +35% | 0 | 0 | — |
case-22 | pass→pass | 14,919 | 11,315 | -24% | 1 | 1 | 0% | 2,300 | 3,305 | +44% | 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 21 counted toward the lift figure. The other 1 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 +27 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is 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.