Version: v1.6.0
codegraph callers <sym> and codegraph node <sym> cap the caller list at the default --limit 20 with no indication the result was truncated — the header shows the displayed count, never "…N more". impact (uncapped) and node's callee trail (+N more) don't have this; the callers list and the JSON / MCP output do.
Repro (a high-fan-in symbol, e.g. a widely-used hook):
codegraph callers useAuth → 20 resultscodegraph callers useAuth --limit 500 → 174 results- The output/header gives no hint that 154 were dropped.
Expected: when truncated, print something like …154 more (raise --limit), or don't cap silently.
Impact: consumers under-count and trust the result → wrong "who calls this / is this dead code?" conclusions. It's especially harmful when an AI agent drives the tool over MCP, since the tool output inherits the cap with no truncation signal.
Version: v1.6.0
codegraph callers <sym>andcodegraph node <sym>cap the caller list at the default--limit 20with no indication the result was truncated — the header shows the displayed count, never "…N more".impact(uncapped) andnode's callee trail (+N more) don't have this; thecallerslist and the JSON / MCP output do.Repro (a high-fan-in symbol, e.g. a widely-used hook):
codegraph callers useAuth→ 20 resultscodegraph callers useAuth --limit 500→ 174 resultsExpected: when truncated, print something like
…154 more (raise --limit), or don't cap silently.Impact: consumers under-count and trust the result → wrong "who calls this / is this dead code?" conclusions. It's especially harmful when an AI agent drives the tool over MCP, since the tool output inherits the cap with no truncation signal.