[QUESTION] Go stack filtering can return an empty stack — should it keep the first frame like JS?

[debugmcpdev]

Surfaced during the #344 review. GoAdapterPolicy.filterStackFrames (packages/shared/src/interfaces/adapter-policy-go.ts) is a plain frames.filter() with no fallback: a goroutine paused entirely inside /runtime/ or /testing/ frames yields an empty get_stack_trace result.

Policies currently disagree on the all-internal edge case:

• JS: returns [frames[0]] (never empty)
• Java: returns the full unfiltered array (never hides the stack)
• Go, .NET: may return []

Docs now describe this accurately (fixed in #344), but the behavioral question stands: an empty stack is a dead end for an agent — it can't tell "no frames" from "all frames filtered", and includeInternals: true is only discoverable if it already knows the filtering exists. Options:

1. Give Go (and .NET) the JS-style first-frame fallback
2. Adopt Java's return-everything fallback everywhere
3. Keep as-is but annotate the empty result (e.g. a note that N internal frames were hidden and how to reveal them)

Option 3 fits the AX principle of replacing what the agent must know with what it can see in the response.

🤖 Generated with Claude Code