get_blast_radius

Answers the question "what breaks if I change this?" Returns every symbol that could be affected by modifying the target, graded across three confidence tiers so you know what is certain vs speculative.

See the mandatory sequence before editing.

Parameters

Name	Type	Required	Description
symbolId	string	yes	Fully-qualified symbol id (<file>::<name>::<kind>)
confidence	"confirmed" | "likely" | "potential"	no	Restrict results to a single tier
intent	string	no	Keyword filter on impacted symbols (e.g. "test", "adapter")

Confidence tiers

Results are grouped into 3 confidence tiers — see Blast Radius tiers.

Example

{
  "symbolId": "packages/cli/src/adapters/storage/sqlite.ts::SqliteStorageAdapter::class"
}

Filter to high-confidence only:

{
  "symbolId": "packages/cli/src/adapters/storage/sqlite.ts::SqliteStorageAdapter::class",
  "confidence": "confirmed"
}

Response

{
  "symbolId": "packages/cli/src/adapters/storage/sqlite.ts::SqliteStorageAdapter::class",
  "impactScore": 27,
  "directDependentsCount": 6,
  "confirmedCount": 6,
  "likelyCount": 14,
  "potentialCount": 7,
  "overallRiskScore": 0.82,
  "impactedSymbols": [
    {
      "symbolId": "packages/cli/src/index.ts::main::function",
      "name": "main",
      "kind": "function",
      "confidence": "confirmed",
      "riskScore": 0.91
    }
  ],
  "byCluster": {
    "packages/cli/src/adapters/storage": 12,
    "packages/cli/src/core/graph": 3,
    "packages/cli/src/adapters/mcp": 8,
    "packages/cli/src/cli": 4
  },
  "crossClusterEdges": 15,
  "multiClusterHint": "Change impacts 4 clusters — multi-team review recommended.",
  "_meta": { "totalItems": 27, "returnedItems": 27, "truncated": false }
}

If the symbol is missing from the index:

{ "found": false, "hint": "Symbol not found. Run \"ctxo index\" to build the codebase index." }

v0.8 cluster fields

Field	When present	Meaning
byCluster	A community snapshot exists (ctxo index after v0.8)	Map of clusterLabel → count of impacted symbols in that cluster
crossClusterEdges	Same	Count of impacted-set edges that cross the target's own cluster
multiClusterHint	byCluster touches ≥ 3 clusters	Human-readable nudge for multi-team review

Killer example: same impactScore, different risk

Two candidate refactors both report impactScore: 10. Before v0.8, they look identical.

Change	byCluster (v0.8)	Verdict
A — rename a private method	{ "src/billing": 10 }	One cluster, one team; proceed.
B — tweak a shared DTO	{ "src/billing": 4, "src/auth": 3, "src/reporting": 3 } + multiClusterHint	Three clusters, three teams; needs coordination.

The raw numbers are identical. The cluster breakdown is the difference between "go" and "stop".

Interpreting overallRiskScore

Score	Risk level	Action
> 0.7	high	Plan carefully, write tests first, review with a teammate
0.3 - 0.7	medium	Read get_why_context before editing
< 0.3	low	Safe to edit, still verify imports via find_importers if unsure

See it in action

See a worked comparison of blind edits vs blast-radius-aware edits: Blast Radius comparison.

Common pitfalls

Skipping this tool

Editing a symbol without calling get_blast_radius first is the #1 cause of broken dependencies. The static graph catches couplings that grep and manual reading will miss, especially through re-exports and dynamic dispatch.

• Treating potential as noise — the tier comes from git co-change history; two files that always change together are coupled even when the parser sees no edge.
• Ignoring _meta.truncated — large impact sets are capped by CTXO_RESPONSE_LIMIT. If truncated, narrow with confidence or intent.
• Using the symbol name instead of id — always pass the full <file>::<name>::<kind> id. Look it up with search_symbols if you only know the name.

Related tools

• get_why_context — required follow-up before editing
• get_logic_slice — forward deps (what this symbol needs)
• find_importers — direct reverse lookup only
• get_pr_impact — blast radius across every changed symbol in a diff