Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add marcos-sponton/frameworks-as-skills --skill working-backwardsgit clone --depth 1 https://github.com/marcos-sponton/frameworks-as-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/marcos-sponton/frameworks-as-skills/working-backwards)<a href="https://agentmods.dev/skills/marcos-sponton/frameworks-as-skills/working-backwards"><img src="https://agentmods.dev/badge/skills/marcos-sponton/frameworks-as-skills/working-backwards/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/marcos-sponton/frameworks-as-skills/working-backwards"><img src="https://agentmods.dev/badge/skills/marcos-sponton/frameworks-as-skills/working-backwards.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00268 | $0.02936 |
| Opus 5 | $0.00134 | $0.01468 |
| Sonnet 5 | $0.00054 | $0.00587 |
| Haiku 4.5 | $0.00027 | $0.00294 |
Grade A, and why
working-backwards scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 10d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 95 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Working Backwards
Bill Carr and Colin Bryar's operational method for building customer-obsessed products at scale — distilled from the 2021 book Working Backwards: Insights, Stories, and Secrets from Inside Amazon (St. Martin's Press) plus their subsequent teaching at Working Backwards LLC, their firm's blog (active March 2026), and podcast appearances (Bill Carr on Lenny Rachitsky, Nov 2023; both authors on First Round Review; joint appearances since book launch).
This skill helps you work in the method, not just recite its vocabulary. It's opinionated because Carr and Bryar are opinionated: mechanisms produce durable behavior; culture is downstream of them. Adopting the PR/FAQ format without the silent-reading meeting, or naming an "STL" who's actually 30% on the initiative, isn't Working Backwards — it's the thing Carr spends most of his airtime warning against.
When this skill activates
Use this skill when the user is:
- Drafting a PR/FAQ or a six-page narrative memo (theirs or someone else's).
- Preparing a product launch, feature bet, or business-case document.
- Critiquing a PRD, spec, or product doc that reads like a feature announcement rather than a customer-need document.
- Structuring or fixing a product-review meeting (silent reading, 6-pager format, review disposition).
- Setting up input metrics for a team or product — or diagnosing why current metrics aren't driving behavior.
- Structuring a single-threaded team, or explaining why a matrixed team keeps stalling.
- Designing a hiring bar for a critical role (Bar Raiser mechanism).
- Running a Weekly Business Review, Operating Plan cycle, or Correction of Errors post-mortem.
- Trying to import "the Amazon way" into their company and asking what's actually load-bearing.
- Deciding whether a bet is worth weeks of upfront thinking before any code is written.
Do NOT use this skill when:
- The user is pre-product/market-fit and iterating cheaply — a full PR/FAQ is overkill in a search-mode context. Suggest Lean Startup / Continuous Discovery instead. Carr and Bryar have publicly noted they target "Series C+ and public companies" precisely because mechanisms matter more once cost-per-bet is high.
- The user's question is purely about corporate strategy at the "how will we win in this market" altitude — that's Playing to Win / Rumelt territory, not Working Backwards. Point them at the right framework.
- The user is asking for a summary of the book. Give them the book link and don't run the method at them.
- The user wants "an Amazon-style memo template" without any customer problem to actually work through. A template alone isn't the mechanism — the review discipline is.
What ships with it
11 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- evals/evals.json 9.1 KB
- README.md 6.4 KB
- references/applications.md 8.4 KB
- references/author-live-sources.md 8.6 KB
- references/examples.md 12 KB
- references/heuristics.md 15 KB
- references/method.md 15 KB
- references/post-book.md 9.1 KB
- references/prompts.md 8.1 KB
- references/sources.md 8.3 KB
- references/voice-and-tone.md 10 KB
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 10d ago First seen · 95 lines · 0 tokens per session scan A a79e28ff66c0
working-backwards is a skill published in the GitHub repository marcos-sponton/frameworks-as-skills (1 stars, last pushed 1mo ago), licensed MIT. It adds 268 tokens to every session and 2,936 once invoked, about $0.0013 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
acquire-skills
Find and install skills for what you're doing — tech or reasoning gaps. Matches activity and tech context to relevant skills. Use when entering unfamiliar tech, stuck on a reasoning challenge, or no skill fits.
find-skills
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express interest in extending capabilities. This skill should be used when the user is looking for functionality that might exist as an installable skill.
endpoint-discovery
Discovers URLs, endpoints, and paths from web applications using katana, gospider, and gau. Use when needing to enumerate application endpoints, find hidden paths, gather historical URLs, or map web application structure during reconnaissance.
param-miner
Discover hidden HTTP parameters in web applications that could reveal hidden functionality or vulnerabilities. Use when testing APIs, when looking for hidden parameters like admin or debug, or when the user mentions parameter discovery.
git-workflow
This skill should be used when the user asks to "create git commit", "manage branches", "follow git workflow", "use Conventional Commits", "handle merge conflicts", or asks about git branching strategies, version control best practices, pull request workflows. Provides comprehensive Git workflow guidance for team…
daily-paper-generator
Use when the user asks to generate daily paper digests on a general topic. This skill supports both arXiv and bioRxiv (or either one), then produces structured Chinese/English summaries for selected papers.