<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Bayarkhuu Tsoggerel</title>
    <description>The latest articles on DEV Community by Bayarkhuu Tsoggerel (@bayarkhuu_tsoggerel).</description>
    <link>https://dev.to/bayarkhuu_tsoggerel</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4201472%2F653bb333-9329-480a-934d-de801779c61e.jpg</url>
      <title>DEV Community: Bayarkhuu Tsoggerel</title>
      <link>https://dev.to/bayarkhuu_tsoggerel</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/bayarkhuu_tsoggerel"/>
    <language>en</language>
    <item>
      <title>Login endpoints with no attempt limit: what coding agents skip and how to catch it</title>
      <dc:creator>Bayarkhuu Tsoggerel</dc:creator>
      <pubDate>Tue, 17 Nov 2026 00:10:19 +0000</pubDate>
      <link>https://dev.to/bayarkhuu_tsoggerel/login-endpoints-with-no-attempt-limit-what-coding-agents-skip-and-how-to-catch-it-326o</link>
      <guid>https://dev.to/bayarkhuu_tsoggerel/login-endpoints-with-no-attempt-limit-what-coding-agents-skip-and-how-to-catch-it-326o</guid>
      <description>&lt;p&gt;Unlimited password guessing in agent-written backends usually comes down to a login route that does exactly what the prompt asked. It looks up the email, compares the hash, returns a token, and passes every test you thought to write. Nothing in it stops someone from guessing passwords against one account as fast as the server answers.&lt;/p&gt;

&lt;h2&gt;
  
  
  The handler that looks finished
&lt;/h2&gt;

&lt;p&gt;Here's a typical login handler (Express, trimmed). As far as it goes, it's correct:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="n"&gt;app&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;post&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"/login"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="k"&gt;async&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;res&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;email&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;password&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;body&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;user&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="n"&gt;db&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;users&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;findByEmail&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;email&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;!&lt;/span&gt;&lt;span class="n"&gt;user&lt;/span&gt; &lt;span class="o"&gt;||&lt;/span&gt; &lt;span class="o"&gt;!&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="k"&gt;await&lt;/span&gt; &lt;span class="n"&gt;bcrypt&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;compare&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;password&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;user&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;passwordHash&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="n"&gt;res&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;status&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="mi"&gt;401&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;json&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;error&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="s2"&gt;"Invalid credentials"&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="p"&gt;}&lt;/span&gt;
  &lt;span class="n"&gt;res&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;json&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;token&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nf"&gt;signToken&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;user&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;id&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The happy path works, the wrong-password path returns a 401, and a reviewer skimming the diff sees clean code. The bug is code that was never written, and a diff only shows code that exists.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why agents skip it
&lt;/h2&gt;

&lt;p&gt;The prompt said "add login", so the agent added login. A rate limit sits outside the feature as described. No test fails without it and no status code looks wrong, so nothing in the agent's own loop tells it something is missing.&lt;/p&gt;

&lt;p&gt;In the backendskills author's bench (15 backend tasks, Sonnet 5.5 and Haiku 5.5, 5 runs per task, scored by a hidden check the agent never sees), unlimited password guessing counts as a critical mistake. Two tasks, login-rate-limit and rename-column, accounted for 20 of the 25 critical mistakes in runs without the skills. 10 of the 15 tasks had no critical mistake either way. The failures cluster in a few places, and login is one of them.&lt;/p&gt;

&lt;h2&gt;
  
  
  How to catch it in review
&lt;/h2&gt;

&lt;p&gt;Start with a test that fails today. Send a run of wrong passwords for one email and assert that you eventually get a 429, then do the same from one IP across many emails. If the endpoint keeps answering 401 forever, there's no limit, whatever the code looks like.&lt;/p&gt;

&lt;p&gt;Then add the limit. It takes two limiters on the route, one keyed on the account and one on the IP. With express-rate-limit (v8):&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight javascript"&gt;&lt;code&gt;&lt;span class="k"&gt;import&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="n"&gt;rateLimit&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;ipKeyGenerator&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt; &lt;span class="k"&gt;from&lt;/span&gt; &lt;span class="s2"&gt;"express-rate-limit"&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

&lt;span class="c1"&gt;// Fail at startup instead of running with NaN if a variable is unset.&lt;/span&gt;
&lt;span class="kd"&gt;function&lt;/span&gt; &lt;span class="nf"&gt;requiredInt&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;name&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;value&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nb"&gt;Number&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nb"&gt;process&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;env&lt;/span&gt;&lt;span class="p"&gt;[&lt;/span&gt;&lt;span class="n"&gt;name&lt;/span&gt;&lt;span class="p"&gt;]&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="k"&gt;if&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="o"&gt;!&lt;/span&gt;&lt;span class="nb"&gt;Number&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;isInteger&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;value&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;||&lt;/span&gt; &lt;span class="n"&gt;value&lt;/span&gt; &lt;span class="o"&gt;&amp;lt;=&lt;/span&gt; &lt;span class="mi"&gt;0&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt;
    &lt;span class="k"&gt;throw&lt;/span&gt; &lt;span class="k"&gt;new&lt;/span&gt; &lt;span class="nb"&gt;Error&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="sb"&gt;`${name} must be set to a positive integer`&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
  &lt;span class="p"&gt;}&lt;/span&gt;
  &lt;span class="k"&gt;return&lt;/span&gt; &lt;span class="n"&gt;value&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;

&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;windowMs&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;requiredInt&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"LOGIN_WINDOW_MS"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;perAccount&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;rateLimit&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="n"&gt;windowMs&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="n"&gt;limit&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nf"&gt;requiredInt&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"LOGIN_MAX_PER_ACCOUNT"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="n"&gt;keyGenerator&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="nb"&gt;String&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;body&lt;/span&gt;&lt;span class="o"&gt;?&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;email&lt;/span&gt; &lt;span class="o"&gt;??&lt;/span&gt; &lt;span class="s2"&gt;""&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;trim&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;toLowerCase&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="n"&gt;standardHeaders&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

&lt;span class="kd"&gt;const&lt;/span&gt; &lt;span class="n"&gt;perIp&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;rateLimit&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="p"&gt;{&lt;/span&gt;
  &lt;span class="n"&gt;windowMs&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="n"&gt;limit&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="nf"&gt;requiredInt&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"LOGIN_MAX_PER_IP"&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="n"&gt;keyGenerator&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt; &lt;span class="o"&gt;=&amp;gt;&lt;/span&gt; &lt;span class="nf"&gt;ipKeyGenerator&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="n"&gt;req&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="n"&gt;ip&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
  &lt;span class="n"&gt;standardHeaders&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kc"&gt;true&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt;
&lt;span class="p"&gt;}&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;

&lt;span class="n"&gt;app&lt;/span&gt;&lt;span class="p"&gt;.&lt;/span&gt;&lt;span class="nf"&gt;post&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="s2"&gt;"/login"&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;perIp&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;perAccount&lt;/span&gt;&lt;span class="p"&gt;,&lt;/span&gt; &lt;span class="n"&gt;loginHandler&lt;/span&gt;&lt;span class="p"&gt;)&lt;/span&gt;&lt;span class="p"&gt;;&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;A few things to check while you're there:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;What the key is.&lt;/strong&gt; You need both limits. A single combined key like &lt;code&gt;${req.ip}:${email}&lt;/code&gt; gives every IP its own budget for every account, so an attacker spreading guesses for one account across many addresses gets through, and so does one address trying a common password against many accounts. The per-account limiter stops the first, and the per-IP limiter stops the second. &lt;code&gt;ipKeyGenerator&lt;/code&gt; groups IPv6 addresses by subnet, so one user can't rotate through their own block of addresses to reset the count. The per-account limit also lets anyone block a user's logins for one window, so keep that window short.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Where the counter lives.&lt;/strong&gt; The default store is in memory, so with more than one app instance each one counts separately. Use a shared store such as Redis in production, with a separate prefix for each limiter.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The other costly routes.&lt;/strong&gt; Password reset, signup, and anything that sends an email or SMS deserve the same treatment.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Pick window and attempt values that fit your users. The test is what keeps the limit from disappearing in the next refactor.&lt;/p&gt;

&lt;h2&gt;
  
  
  Getting the agent to write it the first time
&lt;/h2&gt;

&lt;p&gt;Review catches it, but it's cheaper if the agent writes the limiter up front. backendskills is an MIT-licensed set of three agent skills: api-endpoint, data-integrity and db-migration. api-endpoint tells the agent to scope every query to the caller, turn bad input into a 4xx, and rate-limit login and costly routes.&lt;/p&gt;

&lt;p&gt;In the same bench, critical mistakes went from 25/150 runs without the skills to 2/150 with them (Sonnet 5.5 from 13/75 to 0/75, Haiku 5.5 from 12/75 to 2/75). Runs with the skills cost more and take longer: $0.11 vs $0.06 per run and 23 vs 14 turns on average. 25 runs were excluded because the bench account hit its usage limit.&lt;/p&gt;

&lt;p&gt;It's one author's bench on one sample app, so treat it as a signal and check it against your own tasks. The raw runs, tasks and rerun instructions are in bench/. Each skill also has a "Make it yours" section for your team's conventions, which says "not set" as shipped. If an agent still gets something wrong with the skills loaded, the project asks for an issue with the prompt and the diff.&lt;/p&gt;

&lt;p&gt;To try it on your next backend task, run &lt;code&gt;npx skills add globalmpc/backendskills&lt;/code&gt;. For Codex and other AGENTS.md agents, append dist/AGENTS.md; for Cursor, copy dist/cursor/ into .cursor/rules/.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://autopromo-mu.vercel.app/go/s6cg2ulk" rel="noopener noreferrer"&gt;https://autopromo-mu.vercel.app/go/s6cg2ulk&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Written with help from AI.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>ai</category>
      <category>security</category>
      <category>webdev</category>
      <category>node</category>
    </item>
    <item>
      <title>How MPC plans to connect mining data to on-chain records</title>
      <dc:creator>Bayarkhuu Tsoggerel</dc:creator>
      <pubDate>Sat, 14 Nov 2026 03:03:55 +0000</pubDate>
      <link>https://dev.to/bayarkhuu_tsoggerel/how-mpc-plans-to-connect-mining-data-to-on-chain-records-1436</link>
      <guid>https://dev.to/bayarkhuu_tsoggerel/how-mpc-plans-to-connect-mining-data-to-on-chain-records-1436</guid>
      <description>&lt;p&gt;MPC (Mineral Project Coin) is building an infrastructure layer for mining and commodity real-world assets. I'm part of the MPC team, and this post walks through the design as published today and what has shipped so far.&lt;/p&gt;

&lt;h2&gt;
  
  
  The idea
&lt;/h2&gt;

&lt;p&gt;MPC links real mining data, professional verification and structured digital records so that institutions can review mining assets and prepare them for on-chain issuance. The team describes it as a single rail "from the ground to the record."&lt;/p&gt;

&lt;h2&gt;
  
  
  Five layers
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;Real-World: the mining project and its source data.&lt;/li&gt;
&lt;li&gt;Verification: CPs, experts, labs and AI Assist.&lt;/li&gt;
&lt;li&gt;Data &amp;amp; Oracle.&lt;/li&gt;
&lt;li&gt;Asset: the asset structure, smart contract and registry.&lt;/li&gt;
&lt;li&gt;Application: an explorer, a dashboard and an API.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;On chain, MPC is designed as an RWA issuance layer on BNB Smart Chain, an EVM-compatible chain.&lt;/p&gt;

&lt;h2&gt;
  
  
  What exists today
&lt;/h2&gt;

&lt;p&gt;The platform is in the design and build stage. MPC's own verification and oracle layer is still in design, and the dApp is in development. The part that is live is MPC Unity, the community app.&lt;/p&gt;

&lt;h2&gt;
  
  
  Reading further
&lt;/h2&gt;

&lt;p&gt;The MPC whitepaper is published chapter by chapter as web pages in the MPC docs. Official MPC channels are the website, the MPC Unity app, X, Telegram, Medium, LinkedIn and the contact email; GitHub, npm and pub.dev publish code only and aren't used for announcements. MPC will never contact you unsolicited asking for funds, signatures or credentials.&lt;/p&gt;

&lt;p&gt;To follow the project as it's built, join MPC Unity.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://autopromo-mu.vercel.app/go/2ijrz2wq" rel="noopener noreferrer"&gt;https://autopromo-mu.vercel.app/go/2ijrz2wq&lt;/a&gt;&lt;/p&gt;

</description>
      <category>web3</category>
    </item>
    <item>
      <title>Installing one set of backend rules in Claude Code, Codex and Cursor</title>
      <dc:creator>Bayarkhuu Tsoggerel</dc:creator>
      <pubDate>Sat, 14 Nov 2026 02:54:51 +0000</pubDate>
      <link>https://dev.to/bayarkhuu_tsoggerel/installing-one-set-of-backend-rules-in-claude-code-codex-and-cursor-482k</link>
      <guid>https://dev.to/bayarkhuu_tsoggerel/installing-one-set-of-backend-rules-in-claude-code-codex-and-cursor-482k</guid>
      <description>&lt;p&gt;Our team built backendskills, an MIT-licensed set of agent skills. The goal is for Claude Code, Codex and Cursor to write backend code the way a senior engineer would, safe by default on auth, data, money and schema changes. Below is what the three skills ask of the agent and how to install them in each tool.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the skills cover
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;api-endpoint&lt;/strong&gt; scopes every query to the caller, turns bad input into a 4xx, and rate-limits login and costly routes.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;data-integrity&lt;/strong&gt; uses locks or conditional updates for read-then-write, UNIQUE constraints, and idempotency keys on anything that charges or sends.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;db-migration&lt;/strong&gt; handles schema changes with expand then contract, concurrent indexes, lock timeouts, batched backfills and down migrations.&lt;/p&gt;

&lt;h2&gt;
  
  
  Installing in each agent
&lt;/h2&gt;

&lt;p&gt;Claude Code:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;npx skills add globalmpc/backendskills
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Codex and other agents that read AGENTS.md: append &lt;code&gt;dist/AGENTS.md&lt;/code&gt;.&lt;/p&gt;

&lt;p&gt;Cursor: copy &lt;code&gt;dist/cursor/&lt;/code&gt; into &lt;code&gt;.cursor/rules/&lt;/code&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Changing the rules
&lt;/h2&gt;

&lt;p&gt;The sources are the files in &lt;code&gt;skills/&lt;/code&gt;. To contribute, edit them and run &lt;code&gt;node scripts/build-dist.mjs&lt;/code&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Where it stops
&lt;/h2&gt;

&lt;p&gt;The skills cover endpoints, data integrity and migrations. Other parts of a codebase are out of scope.&lt;/p&gt;

&lt;p&gt;If you run more than one coding agent, how do you keep their rules in sync?&lt;/p&gt;

&lt;p&gt;&lt;a href="https://autopromo-mu.vercel.app/go/1ux9b2lb" rel="noopener noreferrer"&gt;https://autopromo-mu.vercel.app/go/1ux9b2lb&lt;/a&gt;&lt;/p&gt;

</description>
      <category>claudecode</category>
    </item>
    <item>
      <title>Claude keeps saying "load-bearing", compaction keeps eating my "stop saying that", so I built a tool with Claude Code that makes the correction stick</title>
      <dc:creator>Bayarkhuu Tsoggerel</dc:creator>
      <pubDate>Fri, 30 Oct 2026 07:33:09 +0000</pubDate>
      <link>https://dev.to/bayarkhuu_tsoggerel/claude-keeps-saying-load-bearing-compaction-keeps-eating-my-stop-saying-that-so-i-built-a-18pj</link>
      <guid>https://dev.to/bayarkhuu_tsoggerel/claude-keeps-saying-load-bearing-compaction-keeps-eating-my-stop-saying-that-so-i-built-a-18pj</guid>
      <description>&lt;p&gt;"Stop saying load-bearing." &lt;br&gt;
"Don't end with 'Two things…'." &lt;br&gt;
"Verifiable, not falsifiable."&lt;br&gt;
If you use Claude Code, you've probably typed one of these more than once. They're all real complaints in the Claude Code issue tracker: "load-bearing" comes up in 220 issues filed since July (#53454 has 185 reactions), "Two things…" endings are #90799, "falsifiable" is #97305.&lt;br&gt;
The problem is where you type them. Anything you say in chat is temporary: compaction summarises it, and a new session forgets it. Earlier this month someone had to rebuild their instructions by hand from transcripts after idle compaction (#98747, since fixed). Only CLAUDE.md survives.&lt;br&gt;
So our team built nomore, a small CLI that finds the instructions you keep repeating and helps you turn them into CLAUDE.md rules.&lt;/p&gt;

&lt;p&gt;&lt;code&gt;npx @globalmpc/nomore&lt;/code&gt; scans your local Claude Code transcripts and shows repeated instructions, how often and when you last said them. It also flags rules already in your CLAUDE.md that Claude keeps ignoring.&lt;br&gt;
&lt;code&gt;npx @globalmpc/nomore --apply&lt;/code&gt; goes through the suggestions one by one: yes, no or edit. Only the ones you approve get written.&lt;br&gt;
Everything runs locally. No uploads, no LLM by default.&lt;br&gt;
The limitation: not every repeat is a rule. On our own sessions, many were ordinary tasks like "commit" or "continue". That's why nothing is written without your yes.&lt;br&gt;
How Claude Code built it: it wrote 33 of 35 commits (TypeScript, 125 tests), working from a project CLAUDE.md we gave it (tests before "done", a regression test for every bug fix), and we reviewed every step. It made real mistakes: its first grouping threshold lumped 1,489 unrelated issues into one group, which we only caught by checking the actual output.&lt;br&gt;
There's also a starter pack of 20 CLAUDE.md rules drawn from nearly 10,000 public Claude Code issues, each linked to its source.&lt;br&gt;
Free and MIT-licensed: &lt;a href="https://github.com/globalmpc/nomore" rel="noopener noreferrer"&gt;https://github.com/globalmpc/nomore&lt;/a&gt;&lt;br&gt;
What instruction do you find yourself giving Claude Code over and over?&lt;/p&gt;

</description>
      <category>claude</category>
      <category>vibecoding</category>
      <category>ai</category>
      <category>cli</category>
    </item>
    <item>
      <title>We read 10,000 Claude Code issues. Here's what everyone keeps telling it.</title>
      <dc:creator>Bayarkhuu Tsoggerel</dc:creator>
      <pubDate>Mon, 26 Oct 2026 06:15:08 +0000</pubDate>
      <link>https://dev.to/bayarkhuu_tsoggerel/we-read-10000-claude-code-issues-heres-what-everyone-keeps-telling-it-2h5n</link>
      <guid>https://dev.to/bayarkhuu_tsoggerel/we-read-10000-claude-code-issues-heres-what-everyone-keeps-telling-it-2h5n</guid>
      <description>&lt;p&gt;Every Claude Code user has a list of things they've typed more than once. "Don't add comments." "Use pnpm." "Stop apologizing." Each correction fixes one session and then disappears.&lt;/p&gt;

&lt;p&gt;We wanted to know which corrections are universal, so we read the issue tracker.&lt;/p&gt;

&lt;h2&gt;
  
  
  How
&lt;/h2&gt;

&lt;p&gt;We fetched every issue in &lt;code&gt;anthropics/claude-code&lt;/code&gt; labelled &lt;code&gt;area:model&lt;/code&gt; or &lt;code&gt;data-loss&lt;/code&gt; (nearly 10,000), dropped pull requests, duplicates and withdrawn reports, grouped issues with similar titles, and ranked the groups by 1 + reactions per issue. Then we picked, by hand, the 20 behaviours a CLAUDE.md rule can actually change, using only issues that are still open or closed as "not planned". The fetch and mining scripts are in the repo, so the numbers are reproducible.&lt;/p&gt;

&lt;h2&gt;
  
  
  What people complain about most
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fv7ulgktphu9e3w8rrben.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fv7ulgktphu9e3w8rrben.png" alt="Top 10 Claude Code behaviours by reactions" width="800" height="500"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Behaviour&lt;/th&gt;
&lt;th&gt;Reactions&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Adds code comments, even when told to stop&lt;/td&gt;
&lt;td&gt;252&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Can't stop saying "load-bearing"&lt;/td&gt;
&lt;td&gt;185&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Runs destructive git commands&lt;/td&gt;
&lt;td&gt;121&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Replaces curly quotes and em dashes with ASCII&lt;/td&gt;
&lt;td&gt;61&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Says "belt-and-suspenders" everywhere&lt;/td&gt;
&lt;td&gt;43&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Loops when exploring instead of asking&lt;/td&gt;
&lt;td&gt;38&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Edits files in plan mode&lt;/td&gt;
&lt;td&gt;38&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Uses &lt;code&gt;$(...)&lt;/code&gt; and triggers approval prompts&lt;/td&gt;
&lt;td&gt;33&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Drifts into English after reading English output&lt;/td&gt;
&lt;td&gt;30&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Passes &lt;code&gt;rg -r&lt;/code&gt; (that's &lt;code&gt;--replace&lt;/code&gt;)&lt;/td&gt;
&lt;td&gt;29&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;A caveat: each rule currently rests on one issue, so "most-reported" really means "most-reacted". Reactions are a rough signal, but they're public and anyone can check them.&lt;/p&gt;

&lt;h2&gt;
  
  
  How Claude Code helped build it
&lt;/h2&gt;

&lt;p&gt;Claude Code wrote almost all of the code (33 of 35 commits): the script that fetches the issues with the &lt;code&gt;gh&lt;/code&gt; CLI, the title cleanup, the similarity grouping and the ranking, plus 125 tests. We gave it a CLAUDE.md for the project (run the tests before saying done, a regression test for every bug fix) and reviewed each step.&lt;/p&gt;

&lt;p&gt;It also got things wrong. Its first grouping threshold lumped 1,489 issues into a single group; we caught that in the output, tightened it, and the largest group is now 24. The final 20 rules were picked by a person, not the model.&lt;/p&gt;

&lt;h2&gt;
  
  
  The rule pack
&lt;/h2&gt;

&lt;p&gt;It's free and MIT. One command, every rule linked to its issue:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;curl &lt;span class="nt"&gt;-fsSL&lt;/span&gt; https://raw.githubusercontent.com/globalmpc/nomore/main/rules/CLAUDE.md &lt;span class="nt"&gt;-o&lt;/span&gt; nomore-rules.md
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Then add &lt;code&gt;@nomore-rules.md&lt;/code&gt; to your CLAUDE.md (or put both in &lt;code&gt;~/.claude/&lt;/code&gt; for every project).&lt;/p&gt;

&lt;p&gt;The universal list is half the story. The other half is what &lt;em&gt;you&lt;/em&gt; keep repeating. The repo also has a local CLI (preview) that reads your Claude Code transcripts on your machine, with no network, and suggests the corrections you repeat as rules you choose. What surprised us on our own sessions: most repeats were tasks ("commit", "continue"), not preferences, so it shows a short list and you pick.&lt;/p&gt;

&lt;p&gt;Repo: &lt;a href="https://github.com/globalmpc/nomore" rel="noopener noreferrer"&gt;https://github.com/globalmpc/nomore&lt;/a&gt;. Which rule would you add?&lt;/p&gt;

</description>
      <category>claudecode</category>
      <category>ai</category>
      <category>programming</category>
    </item>
  </channel>
</rss>
