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 jpeslar1/john-peslar-ai-skills --skill paul-graham-voicegit clone --depth 1 https://github.com/jpeslar1/john-peslar-ai-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/jpeslar1/john-peslar-ai-skills/paul-graham-voice)<a href="https://agentmods.dev/skills/jpeslar1/john-peslar-ai-skills/paul-graham-voice"><img src="https://agentmods.dev/badge/skills/jpeslar1/john-peslar-ai-skills/paul-graham-voice/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/jpeslar1/john-peslar-ai-skills/paul-graham-voice"><img src="https://agentmods.dev/badge/skills/jpeslar1/john-peslar-ai-skills/paul-graham-voice.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 | $0.00248 | $0.03854 |
| Opus 5.5 | $0.00099 | $0.01542 |
| Sonnet 5.5 | $0.00050 | $0.00771 |
| Haiku 4.5 | $0.00025 | $0.00385 |
Grade A, and why
paul-graham-voice 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 17d 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 — 171 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Paul Graham Voice DNA
Paul Graham is the plain-prose essayist of startups: the Y Combinator co-founder whose essays open with a month-and-year dateline and a thesis in the first sentence, argue in short common words with contractions, and end on a line that reads like a proverb. The signature is calibrated certainty: he tells you how sure he is ("as far as I know," "There are exceptions, but remarkably few") in the same breath as the claim, then pays for the claim with a named company, a named person, or a small piece of arithmetic. No heat, no hype, no lists. Ground every output in references/voice-corpus.md (verbatim excerpts) and references/source-map.md (sources and refresh notes).
Voice DNA
Cadence and rhythm
- Sentences of about 15 words in a narrow band; few under 6, few over 30, almost no fragments: "To write well you have to think clearly, and thinking clearly is hard."
- Paragraphs of 3-6 sentences that chain if/then and turn on "But," "And yet," "Actually," "So": "They meant well, but this is rarely true. If your product is only half as good as your competitor's, you don't get half as many customers. You get no customers, and you go out of business."
- The thesis arrives in sentence one or two, right after the dateline; no anecdotal warm-up unless the anecdote is the argument (Founder Mode opens on Brian Chesky's talk and pivots in four sentences).
- Contractions throughout (you'll, don't, it's, there's, won't); emphasis comes from sentence position, never from bold or italics.
- Spoken (The Social Radars 2026, documented ASR): shorter sentences with self-corrections and tag questions ("right?", "you know?") the essays never use, then a dry one-line close: "It probably set back hairstyles in Silicon Valley by 10 years."
- Native length: a 1,500-4,000 word essay with a Notes section of numbered footnotes and a thanks line; the Hacker News comment is the same voice at 60-120 words.
Hook patterns
- Dateline, then the thesis - "Actually startups take off because the founders make them take off." Deploy as the first sentence of any essay; the word "Actually" signals that a common belief is about to be corrected.
- The received saying, denied - "'You get out,' I heard a thousand times, 'what you put in.' They meant well, but this is rarely true." Deploy to open a contrarian piece; the saying must be one the reader has actually heard.
- The hedged prediction - "I'm usually reluctant to make predictions about technology, but I feel fairly confident about this one: in a couple decades there won't be many people who can write." Deploy once, for the boldest claim in the piece.
- The recipe gate - "The following recipe assumes you're very ambitious." Deploy to open a how-to essay; it filters the reader without flattering them.
- The observation that struck me - "It struck me recently how few of the most successful people I know are mean. There are exceptions, but remarkably few." Deploy for a character or culture essay; the hedge is part of the hook.
- The one-line definition - "A wise person is someone who usually knows the right thing to do." Deploy to open a conceptual essay, then spend the essay testing the definition.
What ships with it
2 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.
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.
- 17d ago Changed · +1 lines cb6c7ebf2543
- 18d ago First seen · 170 lines · 248 tokens per session scan A f2fc50a3e181
paul-graham-voice is a skill published in the GitHub repository jpeslar1/john-peslar-ai-skills (12 stars, last pushed 18d ago), licensed MIT. It adds 248 tokens to every session and 3,854 once invoked, about $0.0010 per session on Opus 5.5. 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 2026-10-19.
paul-graham-voice compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| paul-graham-voicejpeslar1/john-peslar-ai-skills | Skill | 12 | 18d ago | MIT | A | 248 | Claude Code |
| imagegenopenai/codex | Skill | 133k | yesterday | Apache-2.0 | A | 113 | Codex |
| playwright-climicrosoft/playwright | Skill | 99k | yesterday | Apache-2.0 | A | 19 | Claude Code |
| local-ai-agentsmicrosoft/ai-agents-for-beginners | Skill | 79k | 4d ago | MIT | A | 200 | Claude Code, Codex |
| systematic-debuggingobra/superpowers | Skill | 313k | 3d ago | MIT | A | 21 | Claude Code |
| next-cache-components-adoptionvercel/next.js | Skill | 143k | yesterday | MIT | A | 95 | Claude Code |
| next-partial-prefetching-adoptionvercel/next.js | Skill | 143k | yesterday | MIT | A | 103 | Claude Code |