Blast Radius
Find what a code change could break by running real code to verify safety.
Installation
- Make sure Claude is on your device and in your terminal.
Skills load from
~/.claude/skills/when Claude Code starts up — so you need it on your machine first. If you don't have it yet, install it once with the command below, then runclaudein any terminal to verify.One-time setupnpm i -g @anthropic-ai/claude-codeAlready have it? Skip ahead.
- Paste into Claude Code or into your terminal.
This copies the whole skill folder into
~/.claude/skills/blast-radius-savannahfeder/— the SKILL.md plus any scripts, reference docs, or templates the skill ships with. Works for every skill.What this command does
- Makes a new, empty temporary folder.
- Downloads the skill's folder,
.claude/skills/blast-radius, fromgithub.com/savannahfeder/agentbox(branchmain), the GitHub repository it's listed from. You get the version that is there now, not a copy we keep. - Copies that folder to
~/.claude/skills/blast-radius-savannahfeder/. - Deletes the temporary folder.
Installing runs none of the skill's files; it only downloads and copies them. When you later use the skill, Claude follows its SKILL.md and may run scripts that come with it. Read them first.
Faster alternative (instruction-only skills)
Skips the clone and grabs only the SKILL.md file. Don't use this if the skill ships Python scripts, reference markdowns, or asset templates — they won't be downloaded and the skill will fail when it tries to load them.
Quick install (SKILL.md only)Sign up to copy - Restart Claude Code.
Quit and reopen Claude Code (or any other agent that loads from
~/.claude/skills/). New skills are picked up on startup. - Just ask Claude.
Skills auto-activate when your request matches the skill's description — no slash command needed. Trigger phrases live in the skill's own frontmatter; you can read them in the “What this skill does” section above.
Prefer to read the source first? Open on GitHub.
When Claude uses it
Before a change ships, find what it could break outside its own diff, and prove the one fact it is safe because of by running real code rather than writing an argument. Use when a small change touches main/, shared/, the store, worker spawning or anything another part of the app reads, when reviewing a diff you do not trust, or when asked what could this break or what is the blast radius.
What this skill does
Blast radius
Listing the callers is not the job; grep does that in a second. The job is the breakage grep will not show you.
Do not trust your own write-up
A write-up that sounds right reads the same whether or not it is true. So find the one or two facts the change's safety rests on, and push each as far down this list as is cheap. Say where each one stopped.
- You said so. Worth nothing on its own.
- You pointed at the line: a real
file:line, or the library's own source. - You walked the failure step by step and showed it cannot happen.
- You ran it: a script or test that calls the real code and fails loudly if you are wrong.
- You reproduced it in the running app, with the
verify-agentboxskill.
Level 4 is usually one small script in scripts/scratch/ (gitignored) that
imports the module the app ships and calls the exact function you are worried
about. If it should keep holding, make it a test instead.
Where grep stops in this repository
shared/is imported by both the main process and the renderer. A change there has two sets of callers, built two different ways.- Electron reads
main/,shared/andpreload.cjsonce, at boot. A new IPC call from the renderer against an old main process fails silently. Ask whether both halves change together. - Tests that read source as text. Some tests grep a file for a string
instead of importing it.
npm run test:changedfinds them;vitest relateddoes not. - The store is a ledger written by older versions. A renamed or reshaped field has to still read rows written before the change, on every machine that syncs this store, including teammates' through the team cloud.
- How workers start.
sessionArgs, the env scrub, andworker-permissions.jsondecide what every worker can do and who pays for it. A change near them changes all workers at once. - Other machines. Linux paths, Codex or Claude Code installed with npm
under nvm, an app opened from the Dock with a short PATH, a home folder with
a space.
verify-agentboxdrives these. - What the person sees. A rule duplicated between main and renderer, like ranking or inbox membership, drifts. Check there is still one copy.
Steps
- Read the change: the diff, the symbols it adds, changes or deletes, and what it now does differently that the diff does not spell out.
- Find the one fact it is safe because of. Most risky-looking changes are safe because of a single fact, like "this only drops cache entries that are already dead". If that fact holds, most of the risks clear at once. Spend your time here.
- Look where grep stops, using the list above. Read the source of any library you depend on at its pinned version.
- For each risk, give a real chance of it happening and a real cost if it does. Keep the confirmed ones. List the ones you checked and cleared separately. Cite real lines. A search that finds nothing is still an answer; never invent a caller.
- Prove the one fact at level 4 or 5, and paste what happened.
- For a wide change, get a second opinion with the
interrogateskill.
What to hand back
- What it does, including the part that is not obvious.
- The one fact it is safe because of, the level you got it to, and the proof. If you could not prove it, say unproven.
- Risks: how each breaks,
file:line, how likely, how bad, how to check. - Cleared: what you checked and why it is fine.
- Before it ships: the cheapest test that would catch the real bug.
Adapted from pstack's blast-radius skill,
https://github.com/cursor/plugins/tree/main/pstack (MIT).
Related skills
Documentation Co-Authoring
anthropics
Guide structured workflows for writing docs, proposals, and technical specs collaboratively.
MCP Server Builder
anthropics
Build protocol servers that connect language models to external APIs and services.
Skill Builder & Optimizer
anthropics
Create, edit, and optimize Claude skills with performance testing and benchmarking.
Multi-Component Web Artifacts
anthropics
Build complex React artifacts with Tailwind CSS and shadcn/ui components.