<- all tokdocs
2026.10.05@gbtalkstech

Of the Three Claude Code Mods in This Video, Only Blast Radius Is Anthropic's, and the Video Gets It Wrong

AI
Watch on TikTok

View on TikTok ->

A 23.3-second clip (0:23), 720x1280 portrait, HEVC Main video at roughly 239 kbps over 30 fps, HE-AAC audio at 44.1 kHz stereo and roughly 97 kbps, 998,379 bytes on disk (975.0 KiB). Posted 2026-10-05 at 16:22:45 UTC by uploader handle @gbtalkstech, channel nickname "gabe". At capture on 2026-10-06 it held 1,440 views, 60 likes, 1 comment and 4 reposts. The audio track is credited as "original sound" by gabe, which is the creator's own voiceover rather than a licensed song. I read all 12 of the 12 extracted frames and the full 105-word transcript. Frame 1 is a title card reading "CLAUDE CODE MODS JUST DROPPED" in white and orange caps, above three illustrated cartridges shaped like SIM cards or game cartridges, each with a different icon, slotting into a stylized laptop lid with a small orange starburst and a terminal prompt chevron. Frame 2 shows the same three cartridges numbered 1, 2 and 3 in circles, floating above a talking head in a desk chair against a whiteboard with magnetic dots and a framed orange print. Frames 3 through 5 read "1 TOKEN LEDGER" beside a vertical graduated beaker that fills with orange horizontal bars, then caps with a dark red line and a small left-pointing triangle, suggesting a fill threshold. Frames 6 and 7 read "2 CONTEXT VIEW" beside a horizontal strip of four stacked cards labeled, in order, "System", "Files", "Chat" and "Tools" in white sans-serif. Frames 8 and 9 read "3 BLAST RADIUS" above a 6-by-4 grid of 24 document icons, with a glowing orange ring drawn around a cluster of about five and a single solid orange document at the ring's center. Frames 10 through 12 return to the talking head, with frame 11 overlaying "Comment LINK" in white and orange with a down arrow. No screen recording of Claude Code appears at any point. Every visual is motion graphics.

Anthropic really does call them "mods," and the video's launch claim holds

The Whisper transcript opens with "Plot code mods just dropped," which is a transcription error. Frame 1 shows the intended words in on-screen caps: "CLAUDE CODE MODS JUST DROPPED."

The claim itself checks out. "Mod" is Anthropic's own term, not a coinage by the creator and not a third-party marketplace label. Anthropic's mods overview defines it directly: "A mod is a plugin that changes how Claude Code looks and behaves. It's made of JavaScript or TypeScript event handlers." The same page states the rollout: "Mods are on by default. In the terminal, use Claude Code v2.1.287 or later. The Desktop app includes its own copy of Claude Code, and mods work there from v2.1.286." Press coverage dates the launch to October 1, 2026, four days before this video posted, which supports "just dropped."

Mods sit alongside the older extension points rather than replacing them. Settings hooks, Agent Skills, subagents, slash commands, status lines and MCP servers all predate mods and all run from outside the Claude Code process. A mod runs inside it, which is why only a mod can draw a pane or restyle the spinner. The docs add a terminology warning worth repeating: "Claude Code calls both kinds hooks: on these pages, 'hook' means a mod's handler, and the settings-file kind is a 'settings hook'."

Only blast radius is Anthropic's, and the video does not say so

The video presents three mods as one set, in one numbered list, immediately after announcing an Anthropic launch. A viewer would reasonably read all three as first-party.

Anthropic ships exactly three sample mods, in the claude-code/mods directory of the claude-code-playground repository: token-weather, blast-radius and replay-theater. A separate set of built-in mods ships inside Claude Code itself, listed in /plugin under Built-in: cc-plugin-agents-md, cc-plugin-diff, cc-plugin-plugin-authoring, cc-plugin-sec-default, cc-plugin-telemetry and cc-plugin-you-should-know. Source for several of those lives in the mods directory of the Claude Code repository.

Neither list contains a token-ledger or a context-view. Both exist as community mods. token-ledger is published by GitHub user Arunjay4213 in the claude-mods repository. A mod named context-view is published by GitHub user kongyo2 and described as "The context window as one row above the prompt, drawn the way Claude Code draws its own meters." Anthropic's own context-window mod is token-weather, which the docs describe as drawing "a forecast of your context window above the prompt."

The gap matters for support as well as trust. Anthropic says the playground repository "shares them as they are, without support," and a community mod has no support path at all.

The video's description of blast radius is wrong

This is the clearest factual error. The transcript says blast radius "tells you which files are being changed by a current code change." The on-screen graphic reinforces that reading with a grid of file icons and a ring drawn around the ones supposedly in scope.

That is not what the mod does. Anthropic's one-line description is "holds a risky shell command, such as rm -rf or a force push, and shows what it would change, with buttons to proceed or cancel." The mod's README enumerates the specific commands it intercepts: rm -r, rm -f and rm -rf; git reset --hard; git checkout -- . and git restore .; git clean; force pushes including -f, --force-with-lease and +ref syntax; and database migrations via manage.py migrate, db:migrate, alembic upgrade and prisma migrate. For each, the pane shows what the command would destroy, such as the files a delete would remove with their count and total size, or the remote commits a force push would drop. Mechanically, the mod holds a tool.call event until the user presses Proceed or Cancel, and returns { deny } with a reason Claude can act on.

The subject is destructive shell commands that have not run yet, and the output is a permission prompt with evidence attached. Reviewing which files a code edit touched is a different job, covered by Anthropic's replay-theater sample or the built-in /diff pane.

Token ledger reports cost, not how close you are to a session limit

The transcript says token ledger "tells you how many tokens you're using and if you're approaching your session limit." The first half is accurate and the second half belongs to a different mod.

The token-ledger README shows its status line format as $0.4799 session · last turn $0.2394 · 46k in / 5 out · cache 75%, plus a /ledger command that opens a pane of per-turn model, input and output tokens, cache statistics, duration and cost. Those are spend and cache-efficiency figures for the current session. None of them is a plan quota.

Plan limits are tracked by a separate mod in the same repository, quota-meter, which reports the 5-hour and 7-day rolling windows with percent used, reset times and burn-rate projections. The underlying data is available to any mod through the mods API. Anthropic's mods reference documents $.session.usage() as returning { startedAt, context, rateLimits, cost }, where rateLimits is a list of { kind, percentUsed, resetsAt }, and documents a session.measure event that fires "After each turn, and when a plan limit's percent used changes." Two different mods reading two different fields got merged into one sentence.

The security caveat the video omits is the one Anthropic leads with

A 23-second video recommending three installs, with the links moved to a comment, skips the part Anthropic puts in a warning box. The docs state that a mod can "read and write files anywhere your user account can, start programs, and make network requests," can read "environment variables and settings files, including an API key you keep in either," can see "every prompt you send and every tool call Claude makes," and can "approve a tool call before you're asked." The sandboxing note is explicit: "Mods aren't sandboxed. If you turn on sandboxing, the sandbox isolates the Bash commands Claude runs, and a process that a mod starts runs outside it."

There is a documented way to check a mod before loading it. Clone the repository and run claude plugin validate ./some-mod in a shell. The hooks: and calls: lines in the output list every event the mod handles and every mods API call it makes, without executing the mod. That command is the missing step between the comment link and the install.

Key Takeaways

  • Verified: Claude Code mods are a real Anthropic feature, shipped around October 1, 2026, enabled by default from v2.1.287 in the terminal and v2.1.286 in the Desktop app. "Mod" is Anthropic's official term.
  • Verified: blast-radius is one of exactly three sample mods Anthropic publishes in the claude-code-playground repository, alongside token-weather and replay-theater.
  • Correction: Blast radius does not show which files a code change touched. It intercepts destructive shell commands such as rm -rf, git reset --hard, git clean, force pushes and database migrations, shows what each would destroy, and holds the tool.call until you press Proceed or Cancel.
  • Correction: Token ledger and context view are community mods by third-party authors, not Anthropic releases. Presenting all three as one launch set is misleading.
  • Partial correction: Token ledger reports per-turn tokens, cost and cache hit rate. Proximity to a 5-hour or 7-day plan limit is what quota-meter reports, using the rateLimits field of $.session.usage().
  • Unverified: Which specific repositories the creator links in the comments. The post had 1 comment at capture and I did not read comment text, so the context-view and token-ledger projects named here are the closest public matches by name and description, not confirmed as the creator's picks.
  • Context: Anthropic warns that mods run unsandboxed with the user's full permissions and can read secrets and approve tool calls. Run claude plugin validate ./some-mod to list a mod's events and API calls before installing it.

Resources

Published October 5, 2026. Writeup generated from a favorited TikTok.