show-hn-writer — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited show-hn-writer (Agent Skill) and scored it 83/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 4 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 4 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
A bulleted imperative like {match} tells the agent to never reveal, disclose, or mention something to the user. Used adversarially it can instruct the agent to hide its tool calls or lie about what it did — stripping the transparency a user relies on to trust the agent.
A bulleted imperative like {match} tells the agent to never reveal, disclose, or mention something to the user. Used adversarially it can instruct the agent to hide its tool calls or lie about what it did — stripping the transparency a user relies on to trust the agent.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
This skill drafts HN posts using patterns extracted from 250 real top-ranking posts. Every rule below comes from observed data, not convention.
These are the findings from 250 top HN posts scraped April 18 2026. They override any received wisdom about HN writing.
Title length is the single strongest predictor of score.
Default target: under 40 characters. Hard ceiling: 60.
Body text does not affect score. 90% of posts had no body. With-body avg: 189. Without-body avg: 193. Statistically identical. A body is only worth writing if you have genuinely interesting technical detail that won't fit in a title. Never write a body to pad credibility.
Show HN prefix suppresses score. Show HN posts averaged 94 pts vs 186+ for plain statements. The label signals "I want feedback on my thing" which triggers a more skeptical read. Only use "Show HN:" when the project is genuinely novel. Always offer a plain-title alternative.
First-person titles outperform anonymous statements. First-person ("I…", "My…", "We…"): avg 291 pts (n=9). Plain statement: avg 186 pts (n=126). If the builder's perspective is part of the story, lead with it.
Questions generate comments more than upvotes. Question titles avg ratio of comments-to-score above 1.0×. Best for discussions, not for raw score. Ask the user which they're optimising for before writing.
Themes that consistently outperform:
The highest-scoring titles share one trait: they are stories, not topics. "Someone bought 30 WordPress plugins and planted a backdoor in all of them" — 1192 pts. "Google broke its promise to me – now ICE has my data" — 1688 pts. A topic is "WordPress plugin security". A story has an actor, an action, and stakes.
Before drafting, ask:
"Two quick questions:
Do not proceed until you have both answers.
Based on the project and goal, decide which format to use:
Plain title (recommended default) No prefix. Just what it is or what happened. Highest avg score. Use when: sharing a release, article, tool, or event.
Show HN: prefix (use sparingly) Use only when: the project is a working demo, the builder is present to answer questions, and the technical implementation is the interesting part. Avg score is low (94), but it signals authenticity when the project is genuinely novel. Always also draft a plain-title alternative for comparison.
Ask HN: prefix Use when: the goal is discussion, not promotion. Avg engagement ratio > 1.0×. Best for "who is using X?" or "should I do Y?" posts.
Tell HN: prefix Whistleblowing, accountability, or disclosure. One data point at 819 pts. Only use if the post is factual, verifiable, and the builder is named.
The title is the entire post. Treat the body as optional.
Rules derived from data:
it is a literal spec (e.g. "35B-A3B" is a spec, "powerful" is not).
older content that has aged well.
Draft three variants:
Then apply the length test: count chars on each. Flag any over 60.
Ask yourself: does the technical implementation have a detail that cannot fit in the title and that HN engineers would find genuinely interesting?
If yes: write a body (see Step 5). If no: stop at the title. No body is better than a padded body.
The data shows bodies do not increase score. The only reason to write one is if the implementation is interesting enough that engineers will ask "how does this work?" and you want to pre-answer that.
Structure — keep it tight:
Line 1: One sentence. What you built and why. First person. Not "Introducing X." Not "X is a tool that." Just: "I built X because Y."
Lines 2–4: The real reason. Honest. Specific. Was it a problem you hit yourself? Something frustrating at work? A curiosity? "I was annoyed that..." is better than "Developers often struggle with...". The builder's voice is the point.
Lines 5–8: How it actually works. This is what HN reads for. Name the specific technology choices. State the tradeoffs you made and why. One interesting engineering decision is worth more than a feature list.
Line 9: Current state in one sentence. Open source? Free? Alpha? Solo? How long you've been working on it.
Line 10: One closing sentence inviting feedback or questions. "Happy to answer questions about the implementation." / "Criticism welcome." Never ask for upvotes, shares, or sign-ups.
Hard rules:
innovative, best-in-class, streamline, leverage, transform, cutting-edge.
Run through this list. Fix anything that fails before outputting.
Title:
Body (if written):
Post type:
Format exactly as follows. No commentary before or after.
## HN Post
### Recommended title
[title — the shortest, strongest variant]
### Alternative titles
1. [variant 2]
2. [variant 3]
---
### Body
[body text, or "Not recommended — title is sufficient." if Step 4 said no body]
---
### Notes
- Goal: [score / comments] — based on user's answer in Step 1
- Post type used: [plain / Show HN / Ask HN / Tell HN]
- Title length: [N chars]
- Best time to post: Tuesday–Thursday, 8–10 AM US Eastern
- After posting: respond to every comment in the first two hours
- Do not share the link elsewhere for 24 hours — HN penalises vote ringsIf the user wants to check whether similar posts have been submitted recently, or wants to see what is performing in their category right now, run the scraper:
import requests
from concurrent.futures import ThreadPoolExecutor
HN_API = "https://hacker-news.firebaseio.com/v0"
def fetch_item(id_):
try:
r = requests.get(f"{HN_API}/item/{id_}.json", timeout=10)
return r.json() if r.ok else None
except Exception:
return None
ids = requests.get(f"{HN_API}/topstories.json").json()[:250]
with ThreadPoolExecutor(max_workers=20) as ex:
items = [i for i in ex.map(fetch_item, ids) if i]
# Filter by keyword relevant to the user's project
keyword = "YOUR_KEYWORD_HERE"
matches = [i for i in items if keyword.lower() in i.get("title","").lower()]
for i, item in enumerate(matches, 1):
score = item.get("score", 0)
title = item.get("title", "")
by = item.get("by", "")
print(f"{i:>2}. [{score:>4}pts] {title} — {by}")
print(f"\nTotal matching: {len(matches)}")Use the results to:
Results are also appended to hn_log.csv automatically if the full scraper is used.
These are real posts from the top 250. Study the title patterns.
Score | Title ------|------- 1941 | Claude Opus 4.7 1688 | Google broke its promise to me – now ICE has my data 1244 | Qwen3.6-35B-A3B: Agentic coding power, now open to all 1192 | Someone bought 30 WordPress plugins and planted a backdoor in all of them 1141 | DaVinci Resolve – Photo 990 | Codex for almost everything 982 | Stop Flock 909 | A new spam policy for "back button hijacking" 893 | GitHub Stacked PRs 819 | Tell HN: Fiverr left customer files public and searchable 668 | I wrote to Flock's privacy contact to opt out of their domestic spying program 619 | Measuring Claude 4.7's tokenizer costs 561 | God sleeps in the minerals 503 | Want to write a compiler? Just read these two papers (2008)
Best Show HN titles (by score): 341 | Show HN: Smol machines – subsecond coldstart, portable virtual machines 199 | Show HN: PanicLock – Close your MacBook lid disable TouchID → password unlock 187 | Show HN: Every CEO and CFO change at US public companies, live from SEC 177 | Show HN: I made a calculator that works over disjoint sets of intervals 152 | Show HN: MacMind – A transformer neural network in HyperCard on a 1989 Macintosh
Highest comment-to-score ratios (for discussion-optimised posts): 1.91× | Why is IPv6 so complicated? 1.43× | Ask HN: Building a solo business is impossible? 1.20× | Ohio prison inmates built computers and hid them in ceiling 1.13× | Ask HN: Who is using OpenClaw? 1.05× | The future of everything is lies, I guess: Where do we go from here?
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.