blast-radius
Explains the likely impact of a risky shell command before it runs.
Source and capability notes checked · By Anthropic · Updated Oct 2, 2026
What blast-radius looks like in use
No recording yet for blast-radius. The repository README has the details.
Before you install
A mod is not sandboxed. It runs inside Claude Code with your permissions. This one:
- can read Bash commands
- can read file metadata
- can read git state and migration status
Check it yourself before running it: git clone https://github.com/anthropics/claude-code-playground && claude plugin validate ./ · Apache-2.0 licensed
Install
git clone https://github.com/anthropics/claude-code-playground.git
cd claude-code-playground/claude-code/mods && claude plugin marketplace add ./
claude plugin install blast-radius@claude-code-playground-mods --scope user
Needs Claude Code 2.1.287 or later. After installing, run
/reload-plugins if the session was already open.
Full install guide
What changes after installation
Recognized risky commands pause with a report and Proceed or Cancel controls. Examples include recursive removal, hard git resets, clean operations, force pushes and database migrations.
The report can list affected files, disk usage, local changes, remote commits or pending migrations. Cancel denies the original call and tells Claude not to retry automatically.
Good to know
The classifier does not fully parse shell syntax, and only the first recognized risky segment is measured. Proceed runs the entire original line. Use a disposable repository when testing, keep ordinary permission rules enabled and do not treat this official example as comprehensive command security.
Similar mods
secret-redactor
Replaces secrets and personal identifiers with stable placeholders.
collision-guard
Warns when another open Claude chat recently edited the same file.
recording-mode
Masks screen text and blocks selected private files while recording.
budget-guard
Warns near a cost or plan limit and blocks work past it.
Popular mods
agent-profiler
Breaks a session into model wait, tool and subagent time.
Agent Time Machine
Records tool calls and project snapshots on a branching timeline.
context-inspector
Finds repeated reads, similar outputs and compaction-related waste.
loop-detective
Collects evidence when the same Bash failure keeps repeating.