Latest / AI Fire Daily / #652 Neil: Claude Code Mods That Make Your Daily Workflow Easier

Transcript

#652 Neil: Claude Code Mods That Make Your Daily Workflow Easier · AI Fire Daily

• 0:00
  We use advanced AI tools to automate our hardest
• 0:03
  work, yet we still manually babysit token limits
• 0:06
  every single day. Oh, yeah. Right. We track our
• 0:08
  own complex task progress, and we worry about
• 0:11
  two sessions editing the exact same file. It
• 0:14
  is a fascinating modern paradox, really. It really
• 0:16
  is. We built a machine to do the heavy lifting,
• 0:19
  but we still do the administrative chores for
• 0:22
  the machine. Welcome to the deep dive. It's genuinely
• 0:25
  great to be here exploring this today. Today,
• 0:27
  we are exploring a solution to that exact paradox.
• 0:31
  We're looking very closely at Claude code mods.
• 0:34
  Right, and we'll thoroughly explore what these
• 0:36
  tools actually are. Yeah, and then we'll walk
• 0:39
  through a step -by -step building example together.
• 0:41
  We're going to build one entirely from scratch.
• 0:43
  And we're doing this using just plain English.
• 0:46
  I mean, no coding experience is required for
• 0:48
  this at all. Exactly. After that, we'll review
• 0:51
  four specific ready -to -install mods. These
• 0:54
  tools will fundamentally change your daily workflow.
• 0:56
  I'm really excited to get into these specific
• 0:58
  tools. Let's start with the basic foundational
• 1:01
  concept here. We have this paradox of manual
• 1:04
  AI management. How do mods actually solve this
• 1:07
  problem for you? Well, we first need to separate
• 1:10
  plugins from mods. People confuse these two terms
• 1:13
  very frequently. Think of plugins as basic add
• 1:16
  -ons for the system. They give Cloud Code a few
• 1:18
  extra backend capabilities. Okay, let's unpack
• 1:21
  this a little bit further. What's the fundamental
• 1:23
  difference between the two? A plugin might just
• 1:27
  pull data from a server. It doesn't really change
• 1:29
  how you interact with the software. Right. But
• 1:32
  mods go much further than that basic functionality.
• 1:35
  Mods deeply customize how cloud code actually
• 1:38
  works. They change what appears on your screen
• 1:41
  visually. You might want to track your context
• 1:43
  window constantly. You certainly don't want to
• 1:45
  keep checking it manually. Exactly. Or maybe
• 1:48
  you run multiple sessions simultaneously. You
• 1:51
  want a warning before they edit the same file.
• 1:53
  Yeah. Mods let you add these features directly
• 1:56
  to the desktop. And here's where it gets really
• 1:58
  interesting today. You don't need to know how
• 2:00
  to write code. That is the core appeal of this
• 2:02
  entire system. You just describe what you want
• 2:05
  in plain English. Clog code writes the JavaScript
• 2:08
  or TypeScript for you. It feels like stacking
• 2:11
  Lego blocks of data. You're the architect, giving
• 2:13
  the high -level instructions, the AIs, the builders,
• 2:16
  snapping the pieces together. That's a perfect
• 2:18
  way to look at it. The AI handles the underlying
• 2:20
  code architecture completely. You just focus
• 2:23
  on the ultimate workflow goal. The official guide
• 2:26
  says mods run through the plugin system. Right.
• 2:29
  They integrate seamlessly into the official development
• 2:32
  environment. They add custom panels, interactive
• 2:34
  buttons, and visual alerts. Wow. And this all
• 2:37
  happens right inside the Cloud Code interface.
• 2:40
  It allows you to build features fitting your
• 2:42
  exact workflow. You aren't forced into a generic,
• 2:44
  rigid layout anymore. But if we only use plain
• 2:47
  English, what's stopping the AI from just guessing
• 2:49
  the architecture wrong? It relies on the strict
• 2:52
  structure of the official Mods API to keep the
• 2:56
  generated code grounded. So the API acts as a
• 2:59
  rigid blueprint for the plain English instructions?
• 3:02
  Exactly. It prevents the AI from going totally
• 3:04
  off script. It keeps the generated code safe
• 3:07
  and predictable. That makes perfect sense in
• 3:09
  theory. But I kind of want to see how this holds
• 3:11
  up in practice. Of course. The source walked
• 3:14
  through building a specific mod. It was a tool
• 3:16
  called Token Weather. This really demystifies
• 3:18
  the whole building process for you. It shows
• 3:20
  how accessible this technology has actually become.
• 3:23
  Definitely. Token Weather is a custom mod for
• 3:26
  tracking usage. Let's define a crucial term before
• 3:29
  we go any further. Context window. The tool's
• 3:32
  short -term memory for your current chat. Perfect
• 3:35
  definition. So, how do we actually build this
• 3:37
  tracker? If I want to actually build this Token
• 3:40
  Weather tracker, The source suggests starting
• 3:42
  right in the desktop app. No external tools or
• 3:45
  complicated terminal commands are needed. You
• 3:48
  just open the code tab and pick a project folder.
• 3:50
  The source used a folder called marketing -brain
• 3:53
  for this. It feels surprisingly contained and
• 3:56
  straightforward to set up. You just tell Cloud
• 3:58
  Code exactly what you want. You ask it to display
• 4:00
  context usage above the prompt, and you want
• 4:03
  it to update after every single turn. But there's
• 4:06
  a massive pitfall here to avoid. The author of
• 4:09
  the source made this very clear. Yeah, this is
• 4:11
  important. You must tell it to use the official
• 4:13
  mods API. You also have to specifically request
• 4:17
  API hooks. Let's define what those are really
• 4:19
  quickly. API hooks, specific event triggers that
• 4:22
  run code at exact moments. What happens if you
• 4:25
  don't include that specific prompt requirement?
• 4:28
  It just makes a very basic custom status line.
• 4:30
  It won't function as a true visual mod. It'll
• 4:33
  just sit there doing absolutely nothing useful.
• 4:35
  Got it. So you have to mention JavaScript, TypeScript,
• 4:39
  and the plugin structure. You must tell it not
• 4:41
  to modify your status line configuration. Once
• 4:44
  you clarify that, it creates a token weather
• 4:47
  folder. It builds the module and the plugin configuration
• 4:50
  files automatically. It sets up three specific
• 4:53
  hooks to make it work. The first hook is called
• 4:55
  session .start. This checks your context window
• 4:59
  usage when a session begins. The second hook
• 5:01
  is called turn .complete. This updates the usage
• 5:04
  data after each main conversation turn. And the
• 5:07
  final hook is called ui .render. This actually
• 5:10
  displays the token weather bar above your prompt.
• 5:16
  I still wrestle with prompt drift myself. Oh
• 5:18
  man, it happens to absolutely everyone who uses
• 5:21
  these tools. Let's define that term for clarity.
• 5:24
  Prompt drift. Losing track of your original goal
• 5:26
  during long chats. It's so easy to lose track.
• 5:29
  It really is. Having visual token limits on screen
• 5:33
  is incredibly vital. It keeps you completely
• 5:36
  grounded in reality. You know exactly what the
• 5:38
  AI can still remember. So after generating the
• 5:42
  code, you install the mod locally. You do this
• 5:45
  via a local marketplace in the app. You prompt
• 5:48
  it to install the existing token weather mod.
• 5:50
  You tell it to validate the plugin before installing.
• 5:53
  You instruct it to keep the existing files completely
• 5:56
  unchanged. And you do this entirely inside the
• 5:59
  desktop application. No GitHub or external terminals
• 6:03
  required at all. It handles the deployment seamlessly
• 6:06
  in the background. The AI actually created a
• 6:08
  local marketplace called my -mods. It ran the
• 6:11
  validation and started the installation process,
• 6:13
  but there was a slight hiccup during this test.
• 6:16
  Right. Claude initially couldn't find the Claude
• 6:19
  command in its path. Let's define what that means,
• 6:22
  path, the system's map of where software lives.
• 6:25
  Because it couldn't find it, it had to pivot.
• 6:27
  It smartly located the bundled Claude .exe file
• 6:31
  instead. Oh, wow. Yeah, it finished the installation
• 6:33
  job completely on its own. That's a highly impressive
• 6:36
  level of autonomous problem solving. After that,
• 6:39
  you simply reload your Claude code plugins. Right.
• 6:42
  You just type slash reload dash plugins directly
• 6:45
  into the chat window. In the source test, it
• 6:47
  loaded 17 different plugins. It also loaded 158
• 6:50
  skills, 11 agents, and three hooks. That's a
• 6:53
  lot. And you can verify the installation by typing
• 6:56
  forward slash plug -in. Then you get the payoff
• 6:59
  for all this work. You check your brand new token
• 7:02
  weather mod. You send a message and look directly
• 7:04
  above the prompt. And what did it look like?
• 7:07
  The bar showed 23 % context usage immediately.
• 7:11
  It displayed the token count and the recent usage
• 7:13
  change. It even showed the estimated session
• 7:15
  cost right there. Everything is displayed in
• 7:17
  one small convenient visual bar. Yeah, you never
• 7:20
  have to open another clinky menu again. It updates
• 7:23
  automatically after every single conversation
• 7:25
  turn. It's a massive quality of life improvement
• 7:27
  for daily users. Why do we need specific event
• 7:31
  hooks rather than just telling it to update constantly?
• 7:34
  Well, constant updates would drain your system
• 7:35
  resources heavily, whereas hooks tie actions
• 7:38
  to specific necessary moments like a completed
• 7:41
  turn. Right. Hooks guarantee it only processes
• 7:43
  data when something actually changes. Exactly.
• 7:46
  Meteorall sponsor read placeholder. We just walk
• 7:49
  through building a local mod entirely from scratch.
• 7:52
  Now, let's transition to utilizing existing ones
• 7:54
  from a directory. Building is fun, but downloading
• 7:56
  is much faster. Oh, for sure. There's a whole
• 7:59
  ecosystem of these tools out there now. We're
• 8:01
  starting with the most critical workflow issue
• 8:03
  mentioned earlier. We need to deeply discuss
• 8:05
  managing your context limits. This brings us
• 8:07
  directly to the nathercmods marketplace directory.
• 8:11
  There are four specific mods we want to cover
• 8:13
  today. The first one we're looking at is called
• 8:15
  cachekeeper. This is absolutely the best mod
• 8:18
  for long sessions. It solves a massive headache
• 8:21
  for heavy users. It helps you track your context
• 8:24
  usage constantly. It shows how long your prompt
• 8:26
  cache stays active. Let's define what that actually
• 8:29
  means for you listening. Warm cache. Saved memory
• 8:33
  that makes the next AI response cheaper. And
• 8:35
  it estimates your API costs in real time. Installation
• 8:39
  is incredibly simple for this specific tool.
• 8:41
  Yeah, you just visit the cache keeper page and
• 8:43
  copy the instructions. You open Claude Code Desktop
• 8:46
  and start a local session. You just paste those
• 8:49
  copied instructions into the chat window. And
• 8:51
  the AI handles the entire installation process
• 8:54
  for you. Or you can use the graphical interface
• 8:56
  instead. Go to Plugins, then Code, then Nate
• 9:00
  Hurk Mods. Right. You just find CacheKeeper and
• 9:02
  click the plus icon. It really couldn't be any
• 9:05
  easier to do. To use it, you start a new session
• 9:08
  and type slash cache. The source detailed some
• 9:11
  very specific test results here. Typing slash
• 9:14
  cache showed 121k tokens used. It showed a 60
• 9:18
  -minute cache window that was still warm. Usage
• 9:21
  was at 33 % of the five -hour limit. It was also
• 9:25
  at 44 % of the weekly limit. The estimated session
• 9:28
  cost was exactly 65 cents. It gives you complete
• 9:32
  financial transparency for your workflow. It
• 9:34
  even showed a cold send guard enabled at 150k
• 9:38
  tokens. It warns you before your warm cash actually
• 9:41
  expires. That saves you from spending unnecessary
• 9:44
  money on tokens. It prevents those massive unexpected
• 9:48
  billing surprises. There's also a brilliant slash
• 9:51
  handoff feature included here. This is incredibly
• 9:54
  useful for long complex project workflows. But
• 9:58
  is the slash handoff command just doing a bulk
• 10:00
  copy paste of the chat history? No, it actively
• 10:03
  prepares and distills the important context rather
• 10:06
  than just dumping raw text into a new session.
• 10:08
  It saves your core project memory without dragging
• 10:11
  the dead weight along. Exactly. It makes your
• 10:14
  next session fast and highly efficient. It strips
• 10:17
  out the conversational filler and keeps the code.
• 10:20
  Managing the AI's memory is beautifully solved
• 10:21
  by CacheKeeper. Now we must move to managing
• 10:25
  human -level operational issues. We're talking
• 10:27
  about privacy, project scale, and file management.
• 10:31
  These are the things that usually trip developers
• 10:33
  up daily. The second mod to look at is recording
• 10:36
  mode. Privacy is a massive hurdle when sharing
• 10:39
  workspaces. Oh, absolutely. If I'm recording
• 10:42
  a tutorial, my API keys are usually visible.
• 10:45
  Right, and that's exactly what Recording Mode
• 10:47
  is designed to fix. It's for people who record
• 10:49
  tutorials or share screens frequently. It keeps
• 10:52
  sensitive information completely out of public
• 10:54
  view. It hides email addresses, API keys, and
• 10:57
  financial details seamlessly. It places visual
• 11:00
  placeholders directly on your screen, but Cloud
• 11:03
  Code can still use the original information behind
• 11:05
  the scenes. You activate it by simply typing
• 11:08
  slash rec. It instantly redacts everything you
• 11:11
  want hidden from the recording. You can also
• 11:13
  use slash rec strict for stronger protection.
• 11:17
  When you finish recording, you just type slash
• 11:18
  rec off. Always test this with sample data first
• 11:22
  though. It might occasionally miss some highly
• 11:24
  specific sensitive details. Good point. You don't
• 11:27
  want to accidentally leak a production database
• 11:29
  key. You should always check what's visible before
• 11:31
  sharing publicly. The third mod on our list is
• 11:35
  called Goal Meter. This is a progress dashboard
• 11:38
  designed specifically for large projects. It's
• 11:41
  easy to lose track of what Claude has finished,
• 11:43
  especially if you're running several concurrent
• 11:45
  sessions at once. Goal Meter shows the current
• 11:48
  task plan very clearly. It shows completed tasks
• 11:51
  and the overall completion percentage. The source
• 11:54
  tested this with a multi -step workflow task.
• 11:56
  They asked Claude to build a simple landing page.
• 11:59
  Right. It was specifically for an AI newsletter.
• 12:02
  They created a plan, implemented the page, and
• 12:05
  verified it. GoalMeter recognized the workflow
• 12:07
  and tracked the interface progress. It gave a
• 12:10
  visual representation of the entire build process.
• 12:13
  It makes the most sense for those longer complex
• 12:15
  projects. It doesn't add much value for quick
• 12:18
  questions or simple edits. Yeah, that makes sense.
• 12:21
  The fourth and final mod today is Collision Guard.
• 12:24
  This solves a major problem for complex development
• 12:27
  workflows. It really does. Multiple sessions
• 12:29
  might accidentally edit the exact same file.
• 12:32
  They can easily override each other's hard work.
• 12:34
  You could lose hours of prompting in a single
• 12:37
  second. Collision Guard checks a 30 -minute window
• 12:40
  for any recent edits. If it detects a possible
• 12:43
  conflict, it warns you immediately. You activate
• 12:46
  it by using the slash guard command. The source
• 12:49
  tested this on that exact newsletter landing
• 12:51
  page. They asked Claude to create an HTML index
• 12:54
  file. Claude immediately responded that Collision
• 12:57
  Guard was currently turned on. It actively monitored
• 13:00
  plan .md and index .html. It successfully verified
• 13:04
  no conflicts occurred in that 30 minute window.
• 13:08
  Whoa. Imagine scaling to dozens of concurrent
• 13:10
  sessions without the AI ever stepping on its
• 13:13
  own code. That completely changes how we manage
• 13:15
  automated file building. It acts as an incredible
• 13:18
  safety net for developers everywhere. You can
• 13:21
  confidently run parallel tasks without any fear.
• 13:24
  So going back to privacy for a second. Does recording
• 13:26
  mode actually delete your private data from the
• 13:29
  backend logs? No, it only masks the data visually
• 13:32
  on your screen for screen sharing purposes. The
• 13:35
  data remains in the actual workflow. Got it.
• 13:38
  It's a visual shield for your screen, not a data
• 13:41
  eraser. Right. It just keeps your secrets off
• 13:43
  the video recording itself. The AI still uses
• 13:46
  the real data to do the work. Let's step back
• 13:48
  and look at the big idea here. Mods solve workflow
• 13:52
  friction points you might not even realize you
• 13:54
  have. You absolutely don't need to change your
• 13:57
  entire workflow. You don't have to overhaul your
• 13:59
  whole development process today. You just pick
• 14:01
  one single tool to start with today. CashKeeper
• 14:04
  is fantastic for beginners managing those tricky
• 14:07
  token limits. Let plain English do the heavy
• 14:10
  lifting of customizing your environment. You're
• 14:12
  literally talking your custom tools into existence.
• 14:15
  It's a fundamentally new way to interact with
• 14:17
  software. It bridges the gap between idea and
• 14:20
  execution perfectly. It leaves me with a Very
• 14:23
  deep, reflective question. We're already using
• 14:27
  AI to generate these custom tools. The tools
• 14:29
  monitor the AI zone limitations, like tokens.
• 14:32
  They proactively correct the AI zone workflow
• 14:35
  collisions. How long until the system starts
• 14:38
  proactively coding and installing its own mods
• 14:41
  to optimize our workflow before we even ask?
• 14:44
  That's a staggering concept to think about. A
• 14:47
  self -optimizing workspace that evolves alongside
• 14:50
  your habits. Evaluate your own daily pain points
• 14:53
  with AI today. Try installing just one of these
• 14:56
  mods to fix it. Thank you for joining us on the
• 14:58
  Deep Dive.