blast-radius

Blast radius

Ported from the blast-radius skill in cursor/plugins pstack — Cursor, a separate AI-coding-editor product, not to be confused with a text/DB cursor — for use with this harness's Agent tool. Two adaptations from the original:

• Step 6's "run it as an arena" (ask several different vendor models the same question) doesn't map onto this harness's Agent tool, which only spawns Claude-family models. The equivalent here is spawning several independent general-purpose agents, each drawing a distinct model/effort combination from the same four-combination panel how's critique mode uses — see Step 6 below.
• "Write it through unslop" (a skill from that other editor for stripping AI-slop prose) has no port here — if your project enforces an equivalent prose-quality or AI-slop-detection hook, writing the blast-radius writeup will get that scoring for free; otherwise apply the same discipline manually (no narrative comments, no hedge-padding, no filler).

If your environment has code-graph MCP tools (e.g. code-review-graph's get_impact_radius_tool/get_affected_flows_tool/detect_changes_tool, or tokensave's tokensave_callers/ tokensave_impact/tokensave_context), they do the mechanical "find the callers" part of Step 3 far more cheaply than grep — use them there instead of manual search when the target is in an indexed repo. They don't replace Steps 2, 4, and 5, which are the actual point of this skill.

Third adaptation: prefer a code-graph tool over raw grep, when one is connected. The original has Step 3 read library source and grep by hand. If a code-graph MCP indexes the target repo, route through its context/search tools first, the same preference how and why apply. Raw Read/grep stay the fallback for anything the code-graph tool genuinely doesn't index: a pinned dependency's code under node_modules/vendor/site-packages, a file kind outside its indexed languages, or an environment with no such tool connected at all.

Find what a change breaks somewhere else, before it ships. Use for "blast radius of X", "what could this break", or reviewing a small diff you don't trust yet.

Companion to how and why. how tells you what the code does. why tells you why it's shaped that way. Blast radius tells you what it breaks somewhere else.

Listing the callers is not the job. tokensave_callers/tokensave_impact, or code-review-graph's impact-radius tools, can find those in a second. The job is the breakage a caller list won't show you.