<?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: ran guo</title>
    <description>The latest articles on DEV Community by ran guo (@ran_guo_5ed7f647ab1b09c50).</description>
    <link>https://dev.to/ran_guo_5ed7f647ab1b09c50</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%2F4145460%2F92d5bb29-ffc6-4d82-9b61-3acbb3f52f7e.png</url>
      <title>DEV Community: ran guo</title>
      <link>https://dev.to/ran_guo_5ed7f647ab1b09c50</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/ran_guo_5ed7f647ab1b09c50"/>
    <language>en</language>
    <item>
      <title>Six coding agents, one menu bar: where the "waiting on you" signal actually lives</title>
      <dc:creator>ran guo</dc:creator>
      <pubDate>Fri, 30 Oct 2026 06:57:11 +0000</pubDate>
      <link>https://dev.to/ran_guo_5ed7f647ab1b09c50/six-coding-agents-one-menu-bar-where-the-waiting-on-you-signal-actually-lives-58ia</link>
      <guid>https://dev.to/ran_guo_5ed7f647ab1b09c50/six-coding-agents-one-menu-bar-where-the-waiting-on-you-signal-actually-lives-58ia</guid>
      <description>&lt;p&gt;I run six coding agents on the same laptop: Qoder, Claude Code, Codex, ZCode, WorkBuddy, OpenCode. Not for a benchmark — for real work, on different projects, at the same time.&lt;/p&gt;

&lt;p&gt;What I did not expect is that the hard part is not throughput. It is knowing which one is stopped.&lt;/p&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%2Fd53m0y698aysyxlszydp.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%2Fd53m0y698aysyxlszydp.png" alt="Six agents in one menu bar panel, grouped by state" width="800" height="1139"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Three windows, three identical prompts
&lt;/h2&gt;

&lt;p&gt;Last week I spent eleven minutes rewriting a prompt in one window while another session was sitting on a permission question, and a third had already died on an error. All three looked the same on screen: cursor blinking, empty input, exactly how I left them.&lt;/p&gt;

&lt;p&gt;Tiling the windows does not fix this, because a window shows its last render, not its current state. A killed CLI leaves a prompt behind that looks identical to an idle one. You type into it and discover the history is gone.&lt;/p&gt;

&lt;p&gt;Desktop notifications do not fix it either. Notifications carry "task finished". The two states that cost me time are &lt;em&gt;it is asking me something&lt;/em&gt; and &lt;em&gt;it stopped on an error&lt;/em&gt;, and neither of them fires a notification — from the tool's point of view the first one did its job and the second one never finished.&lt;/p&gt;

&lt;h2&gt;
  
  
  The state was already on disk
&lt;/h2&gt;

&lt;p&gt;Every one of those six tools writes a local log: Codex keeps rollouts, Claude Code keeps transcripts, the others keep session stores and local databases. Different formats, same payload: what happened last in this session.&lt;/p&gt;

&lt;p&gt;So I wrote Agent Ping, a menu bar panel that reads those logs and sorts sessions into five states, grouped, with the "waiting on you" group first — and it carries the verbatim question, not "1 pending item". Click a row, you land back in that session: in-app sessions open in their app, CLI sessions resume.&lt;/p&gt;

&lt;p&gt;The unit of value is one row: a session, a state, a way back.&lt;/p&gt;

&lt;h2&gt;
  
  
  Three things it deliberately does not do
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;It does not orchestrate.&lt;/strong&gt; Plenty of tools dispatch work between agents now, and that is a bigger idea than a status panel. But once an orchestrator owns the flow, it also owns the explanation of "what is happening now", and when it breaks you debug the orchestrator instead of your agents. A panel that only reads logs has a cheap property: when the panel says a session is stuck on a permission prompt, you can open the same file and check.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;It makes no network calls.&lt;/strong&gt; The whole binary opens zero connections — &lt;code&gt;nm -u&lt;/code&gt; on the shipped bundle shows no socket symbols. These logs contain your prompts, your code, and the file names you are working on. A tool that sits next to every session all day is the last place that should phone home.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;It does not claim rewards on your behalf.&lt;/strong&gt; Some of these tools keep free-credit state on local disk. The panel shows it and, at most, brings that app to the front. You click.&lt;/p&gt;

&lt;h2&gt;
  
  
  What it cannot do, listed on screen
&lt;/h2&gt;

&lt;p&gt;The first release has specific gaps, and they are in the README rather than in a support thread:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;ZCode does not expose a session-level URL. "Jump to that session" is a workaround that drives its own ⌘K palette and needs Accessibility permission; without it the row degrades to "activate the app". If the workspace was never opened inside ZCode's UI, its search pool does not contain the session at all.&lt;/li&gt;
&lt;li&gt;The Linux status item and banners were verified against a third-party tray host inside a container, not on a real KDE/GNOME desktop. Clicking a banner does not navigate on Linux.&lt;/li&gt;
&lt;li&gt;Six sources, not "all agents". Other formats are not parsed, so those tools do not appear.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Install
&lt;/h2&gt;

&lt;p&gt;macOS 13+ is a menu bar app; Linux is an X11 tray item; six UI languages. There is no prebuilt binary — the formula compiles from the tagged source, about a minute on Apple silicon, and needs Xcode command line tools once.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;brew tap DreamOfXM/agent-ping https://github.com/DreamOfXM/agent-ping
brew trust dreamofxm/agent-ping
brew &lt;span class="nb"&gt;install &lt;/span&gt;agent-ping
agent-ping
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;Repo: &lt;a href="https://github.com/DreamOfXM/agent-ping" rel="noopener noreferrer"&gt;https://github.com/DreamOfXM/agent-ping&lt;/a&gt; (Apache-2.0). Site: &lt;a href="https://dreamofxm.github.io/agent-ping/en/" rel="noopener noreferrer"&gt;https://dreamofxm.github.io/agent-ping/en/&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  When this is useless
&lt;/h2&gt;

&lt;p&gt;If you run one agent, a status panel is overhead — one session cannot be waiting on you out of sight. It starts paying off when two or three run at once and at least one is parked on a confirmation or already dead, and the only way to find out is clicking through windows.&lt;/p&gt;

&lt;p&gt;That is the whole thesis. Everything else in the repo is the part where that turns out to be harder than it sounds: six log formats, six ways of encoding "needs input", and three of them inconsistent about whether a session that errored is still alive.&lt;/p&gt;

</description>
      <category>ai</category>
      <category>productivity</category>
      <category>opensource</category>
      <category>showdev</category>
    </item>
    <item>
      <title>What App Sandbox actually takes away from a Mac disk cleaner</title>
      <dc:creator>ran guo</dc:creator>
      <pubDate>Sun, 27 Sep 2026 12:03:39 +0000</pubDate>
      <link>https://dev.to/ran_guo_5ed7f647ab1b09c50/what-app-sandbox-actually-takes-away-from-a-mac-disk-cleaner-5cmo</link>
      <guid>https://dev.to/ran_guo_5ed7f647ab1b09c50/what-app-sandbox-actually-takes-away-from-a-mac-disk-cleaner-5cmo</guid>
      <description>&lt;p&gt;A disk cleaner is a badly-shaped citizen of the App Sandbox: its whole job is to look at files the user did not explicitly pick. DiskWise ships twice — one build inside the sandbox (Mac App Store), one outside it (direct download) — and the two builds are not the same program. Same binary logic, different capabilities.&lt;/p&gt;

&lt;p&gt;This post is the list of differences, each one pinned to a line in the public repo so you can check it instead of trusting it. If you're planning a sandboxed utility that touches other apps' data, most of these will show up for you too.&lt;/p&gt;

&lt;p&gt;Paths below are relative to the repo root: &lt;a href="https://github.com/DreamOfXM/diskwise" rel="noopener noreferrer"&gt;https://github.com/DreamOfXM/diskwise&lt;/a&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  1. You cannot scan the disk
&lt;/h2&gt;

&lt;p&gt;Not "you shouldn't" — the scope constant literally resolves to a different value per build:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight swift"&gt;&lt;code&gt;&lt;span class="kd"&gt;public&lt;/span&gt; &lt;span class="kd"&gt;static&lt;/span&gt; &lt;span class="k"&gt;var&lt;/span&gt; &lt;span class="nv"&gt;effective&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kt"&gt;ScanScope&lt;/span&gt; &lt;span class="p"&gt;{&lt;/span&gt; &lt;span class="kt"&gt;HomeAccess&lt;/span&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="n"&gt;runsSandboxed&lt;/span&gt; &lt;span class="p"&gt;?&lt;/span&gt; &lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="nv"&gt;user&lt;/span&gt; &lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="n"&gt;disk&lt;/span&gt; &lt;span class="p"&gt;}&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;&lt;code&gt;Sources/DiskCleanerCore/ScanScope.swift:24&lt;/code&gt;&lt;/p&gt;

&lt;p&gt;And the root list follows: the sandboxed build gets the home directory and &lt;code&gt;/Applications&lt;/code&gt;; the direct build additionally appends the system roots (&lt;code&gt;/Library&lt;/code&gt;, &lt;code&gt;/opt&lt;/code&gt;, &lt;code&gt;/private&lt;/code&gt;, &lt;code&gt;/usr/local&lt;/code&gt;, &lt;code&gt;/Volumes&lt;/code&gt;) at &lt;code&gt;ScanScope.swift:79-81&lt;/code&gt;.&lt;/p&gt;

&lt;p&gt;Worth noting &lt;em&gt;why&lt;/em&gt; there's no scope switch in the settings UI any more. There used to be two options, and one machine exposed the bug: the overview ring chart was accounted from the user-scope numbers while the list below was built from full-disk numbers. Two denominators on one screen, neither reconcilable with the other. "The numbers don't match" is not a UX nit for this category — it's the thing that ends trust. So the app only ever scans what it can actually see, and says so.&lt;/p&gt;

&lt;h2&gt;
  
  
  2. Reading &lt;code&gt;~/Library&lt;/code&gt; costs exactly one user gesture, forever
&lt;/h2&gt;

&lt;p&gt;There is no entitlement that means "please let me read the user's Library". &lt;code&gt;com.apple.security.files.user-selected.read-write&lt;/code&gt; means &lt;em&gt;user-selected&lt;/em&gt;. So the app asks the user to select one directory and keeps a security-scoped bookmark for it:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;code&gt;NSOpenPanel&lt;/code&gt; opened with &lt;code&gt;canChooseDirectories = true&lt;/code&gt;, pre-pointed at the real home dir, one prompt labelled "grant access" (&lt;code&gt;Sources/DiskCleaner/HomeGrant.swift:24-36&lt;/code&gt;).&lt;/li&gt;
&lt;li&gt;The returned URL is turned into a bookmark and stored in &lt;code&gt;UserDefaults&lt;/code&gt; (&lt;code&gt;HomeAccess.swift:48&lt;/code&gt;).&lt;/li&gt;
&lt;li&gt;Every launch resolves the bookmark and re-adopts the scope.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Until that grant exists, the app shows an authorization screen and the sidebar is locked — no scan is issued at all (&lt;code&gt;DiskCleanerApp.swift:242&lt;/code&gt;, &lt;code&gt;:309-310&lt;/code&gt;). That's a product decision, not a limitation: a cleaner that silently returns "nothing to clean" because it can't see anything is worse than one that refuses to answer.&lt;/p&gt;

&lt;p&gt;The full entitlement set is three keys, and there is deliberately no &lt;code&gt;com.apple.security.temporary-exception.files.home-relative-path.read-write&lt;/code&gt; list (&lt;code&gt;build_app/entitlements-appstore.plist:9,11,13&lt;/code&gt;). Those temporary exceptions are the classic route for this category and they are also a review red line; the repo's own architecture notes say so (&lt;code&gt;docs/ARCHITECTURE.md:203-205&lt;/code&gt;).&lt;/p&gt;

&lt;h2&gt;
  
  
  3. Inside the sandbox, &lt;code&gt;NSHomeDirectory()&lt;/code&gt; lies to you — quietly
&lt;/h2&gt;

&lt;p&gt;This one would have shipped as a silent bug if it hadn't been caught.&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight swift"&gt;&lt;code&gt;&lt;span class="c1"&gt;// After entering the sandbox, NSHomeDirectory() and homeDirectoryForCurrentUser&lt;/span&gt;
&lt;span class="c1"&gt;// get rewritten to ~/Library/Containers/&amp;lt;bundle id&amp;gt;/Data. Using that as a scan&lt;/span&gt;
&lt;span class="c1"&gt;// root doesn't throw. It scans an empty shell.&lt;/span&gt;
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;(&lt;code&gt;HomeAccess.swift:5-10&lt;/code&gt;)&lt;/p&gt;

&lt;p&gt;The failure mode is the worst available: no error, just "your Mac looks clean". The fix is to get the home directory from the password entry instead of Foundation:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight swift"&gt;&lt;code&gt;&lt;span class="k"&gt;guard&lt;/span&gt; &lt;span class="k"&gt;let&lt;/span&gt; &lt;span class="nv"&gt;pw&lt;/span&gt; &lt;span class="o"&gt;=&lt;/span&gt; &lt;span class="nf"&gt;getpwuid&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nf"&gt;getuid&lt;/span&gt;&lt;span class="p"&gt;())&lt;/span&gt; &lt;span class="k"&gt;else&lt;/span&gt; &lt;span class="p"&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;return&lt;/span&gt; &lt;span class="kt"&gt;URL&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nv"&gt;fileURLWithPath&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="kt"&gt;String&lt;/span&gt;&lt;span class="p"&gt;(&lt;/span&gt;&lt;span class="nv"&gt;cString&lt;/span&gt;&lt;span class="p"&gt;:&lt;/span&gt; &lt;span class="n"&gt;pw&lt;/span&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="n"&gt;pointee&lt;/span&gt;&lt;span class="o"&gt;.&lt;/span&gt;&lt;span class="n"&gt;pw_dir&lt;/span&gt;&lt;span class="p"&gt;),&lt;/span&gt; &lt;span class="nv"&gt;isDirectory&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;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;(&lt;code&gt;HomeAccess.swift:14-17&lt;/code&gt;)&lt;/p&gt;

&lt;h2&gt;
  
  
  4. &lt;code&gt;startAccessingSecurityScopedResource()&lt;/code&gt; returns &lt;code&gt;true&lt;/code&gt; for processes that aren't sandboxed
&lt;/h2&gt;

&lt;p&gt;So it can't be your sandbox test. The app detects the environment at runtime from the tell-tale path instead (&lt;code&gt;HomeAccess.swift:29-31&lt;/code&gt;) — deliberately not with a compile-time flag, because the same binary should behave correctly in both places.&lt;/p&gt;

&lt;p&gt;A second, related trap: the URL handed back by &lt;code&gt;NSOpenPanel&lt;/code&gt; does &lt;strong&gt;not&lt;/strong&gt; work if you call &lt;code&gt;startAccessingSecurityScopedResource()&lt;/code&gt; on it directly — that returns false. You have to write the bookmark, resolve it back into a new URL, and access &lt;em&gt;that&lt;/em&gt; one (&lt;code&gt;HomeAccess.swift:41-47&lt;/code&gt;). Two steps of ceremony to obtain access you'd think the panel just gave you.&lt;/p&gt;

&lt;h2&gt;
  
  
  5. Emptying the Trash is denied — without even showing a prompt
&lt;/h2&gt;

&lt;p&gt;The obvious design is: the app moves files to the Trash, and "empty Trash" is delegated to Finder via one Apple Event. The entitlement is there for it (&lt;code&gt;automation.apple-events&lt;/code&gt;). On macOS 26.6 the sandboxed build finds out that this path doesn't exist:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight plaintext"&gt;&lt;code&gt;// Measured in the sandboxed build (macOS 26.6): the event sent to Finder is killed
// by appleeventsd (`deny appleevent-send com.apple.finder`). No consent dialog is
// ever shown — TCC is never consulted.
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;(&lt;code&gt;Sources/DiskCleaner/Views/TrashView.swift:208-210&lt;/code&gt;)&lt;/p&gt;

&lt;p&gt;Which means the pre-granted &lt;code&gt;automation.apple-events&lt;/code&gt; entitlement buys nothing in the store build: the denial happens upstream of consent. Consequence for the UI: a button that can never light up is removed rather than left as decoration. &lt;code&gt;canCommandFinder = !HomeAccess.runsSandboxed&lt;/code&gt; (&lt;code&gt;TrashView.swift:211&lt;/code&gt;), and the sandboxed build shows "Empty in Finder" (&lt;code&gt;:91&lt;/code&gt;), which opens a Finder window on &lt;code&gt;~/.Trash&lt;/code&gt; and lets the user press ⌘⇧⌫ themselves (&lt;code&gt;:213-216&lt;/code&gt;). The AppleScript that actually empties the trash (&lt;code&gt;Scanner.swift:660-666&lt;/code&gt;) only ever runs in the direct build.&lt;/p&gt;

&lt;p&gt;So "does your app empty my Trash?" has two honest answers, and they depend on where you installed it from. Writing copy for a dual-channel app means writing it per channel.&lt;/p&gt;

&lt;h2&gt;
  
  
  6. No subprocess for the interesting parts
&lt;/h2&gt;

&lt;p&gt;&lt;code&gt;docker system df&lt;/code&gt; is the only way to report Docker's reclaimable space &lt;em&gt;by kind&lt;/em&gt; (images, build cache, volumes). It's a subprocess, and subprocesses are where the two builds split:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Direct build: &lt;code&gt;/bin/sh -c "command -v docker"&lt;/code&gt; (&lt;code&gt;ScanJobs.swift:515&lt;/code&gt;), then &lt;code&gt;docker info&lt;/code&gt; / &lt;code&gt;docker system df&lt;/code&gt; / &lt;code&gt;docker image ls&lt;/code&gt;.&lt;/li&gt;
&lt;li&gt;Sandboxed build: the CLI chain is cut off by &lt;code&gt;if !HomeAccess.runsSandboxed&lt;/code&gt; at &lt;code&gt;ScanJobs.swift:530&lt;/code&gt;. It falls back to enumerating &lt;code&gt;~/Library/Containers/com.docker.docker&lt;/code&gt; (&lt;code&gt;:527&lt;/code&gt;) and reporting one aggregate number.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;That's a real capability difference users can see: virtual disk size is a single blob in the store build and a breakdown in the direct one.&lt;/p&gt;

&lt;p&gt;The only &lt;code&gt;Process()&lt;/code&gt; that survives into the sandboxed build is &lt;code&gt;/usr/sbin/diskutil apfs list -plist&lt;/code&gt; (&lt;code&gt;Scanner.swift:380&lt;/code&gt;, spawned at &lt;code&gt;:519&lt;/code&gt;) — read-only volume accounting. And &lt;code&gt;~/.Trash&lt;/code&gt; itself is unreadable even with the home grant: the doc comment says so (&lt;code&gt;Scanner.swift:624&lt;/code&gt;) and the &lt;code&gt;contentsOfDirectory&lt;/code&gt; guard returns &lt;code&gt;nil&lt;/code&gt;, not zero (&lt;code&gt;:628&lt;/code&gt;). The UI keeps "unknown" and "0 bytes" as different states, because collapsing them is how you accidentally tell someone their Trash is empty.&lt;/p&gt;

&lt;h2&gt;
  
  
  7. What the sandbox gives back
&lt;/h2&gt;

&lt;p&gt;The trade isn't only loss. Being inside the sandbox makes a claim cheap to check and expensive to fake.&lt;/p&gt;

&lt;p&gt;Two years of arguing about cleaner apps all come down to "what does it actually touch, and does it phone home". The sandboxed build's answer is structural: three entitlements, no network client key, no privileged helper. That's a grep, not a promise:&lt;br&gt;
&lt;/p&gt;

&lt;div class="highlight js-code-highlight"&gt;
&lt;pre class="highlight shell"&gt;&lt;code&gt;&lt;span class="nb"&gt;grep&lt;/span&gt; &lt;span class="nt"&gt;-rn&lt;/span&gt; &lt;span class="s2"&gt;"URLSession&lt;/span&gt;&lt;span class="se"&gt;\|&lt;/span&gt;&lt;span class="s2"&gt;NWConnection&lt;/span&gt;&lt;span class="se"&gt;\|&lt;/span&gt;&lt;span class="s2"&gt;import Network&lt;/span&gt;&lt;span class="se"&gt;\|&lt;/span&gt;&lt;span class="s2"&gt;CFNetwork&lt;/span&gt;&lt;span class="se"&gt;\|&lt;/span&gt;&lt;span class="s2"&gt;WebSocket&lt;/span&gt;&lt;span class="se"&gt;\|&lt;/span&gt;&lt;span class="s2"&gt;socket(&lt;/span&gt;&lt;span class="se"&gt;\|&lt;/span&gt;&lt;span class="s2"&gt;getaddrinfo&lt;/span&gt;&lt;span class="se"&gt;\|&lt;/span&gt;&lt;span class="s2"&gt;dataTask"&lt;/span&gt; Sources/
&lt;/code&gt;&lt;/pre&gt;

&lt;/div&gt;



&lt;p&gt;The only hit in the whole source tree is a line in &lt;code&gt;Localizable.strings&lt;/code&gt; — the UI string that says the app has no telemetry. Same for &lt;code&gt;SMAppService&lt;/code&gt;, &lt;code&gt;SMJobBless&lt;/code&gt;, &lt;code&gt;AuthorizationExecuteWithPrivileges&lt;/code&gt;, "with administrator privileges": zero hits, i.e. no helper tool and no admin password prompt anywhere in the product.&lt;/p&gt;

&lt;p&gt;Two more things the architecture was pushed into, which turned out to be worth having:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;One delete path.&lt;/strong&gt; Every deletion in the app funnels into &lt;code&gt;FileManager.trashItem(at:resultingItemURL:)&lt;/code&gt; (&lt;code&gt;Scanner.swift:591-596&lt;/code&gt;). There is no code path that unlinks a file. Auditing "can this destroy data?" is one function, not a search.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Undo that cannot outlive the session.&lt;/strong&gt; Undo is an append-only &lt;code&gt;[TrashRecord]&lt;/code&gt; in memory (&lt;code&gt;DiskCleanerApp.swift:92&lt;/code&gt;, &lt;code&gt;:136&lt;/code&gt;), never persisted. Restoring writes back to the original path and appends a suffix on collision rather than overwriting (&lt;code&gt;Scanner.swift:604-618&lt;/code&gt;). "Undo stops working when you quit" is not a cleanup routine, it's the absence of a storage layer — which is also why the app can't leave a stale database behind.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  The two rules from this
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;
&lt;strong&gt;If a capability differs per channel, it must differ in the UI too.&lt;/strong&gt; A sandboxed build that renders a button it can't execute teaches users that the app lies. Two of the fixes above exist for exactly that reason.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Don't write marketing copy from a list of assumed permissions.&lt;/strong&gt; "It asks for Apple Events, therefore it empties your Trash via Finder" was a sentence that read fine and was false. Every capability sentence needs a measured line number behind it, or it belongs in the "we should test this" column.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Both builds are in the open: &lt;a href="https://github.com/DreamOfXM/diskwise" rel="noopener noreferrer"&gt;https://github.com/DreamOfXM/diskwise&lt;/a&gt; — the store build's &lt;code&gt;ScanScope.swift:24&lt;/code&gt; is a two-way fork you can read in one line, which is more than most dual-channel apps will tell you.&lt;/p&gt;

</description>
      <category>swift</category>
      <category>privacy</category>
      <category>opensource</category>
      <category>programming</category>
    </item>
  </channel>
</rss>
