[IDEA] Evaluate Go/Rust rewrite only if performance becomes a real bottleneck

[luongnv89]

Description

Consider whether agent-skill-manager should eventually be rebuilt (fully or partially) in a more performance-oriented language such as Go or Rust.

Important Constraint

This is not a current priority.

For now, ASM performance is acceptable and the focus should remain on product direction, shipping, and higher-leverage features.

This should only be revisited if we begin to hit real performance bottlenecks that justify the complexity of a rewrite.

Why Keep This Issue

The idea is worth capturing because a rewrite could eventually become relevant if ASM grows into:

• much larger catalogs / indexes
• heavier local databases / sync pipelines
• more complex install/update resolution
• benchmark/evaluation workloads
• higher-performance CLI and backend expectations

Questions for a Future Evaluation

If performance becomes a problem later, evaluate:

• what exactly is slow?
• is the bottleneck CPU, I/O, startup time, memory, or dependency/runtime overhead?
• can the hot paths be optimized without a rewrite?
• would a partial rewrite be enough?
• which language fits better: Go for distribution/CLI ergonomics, or Rust for tighter control/performance?

Acceptance Criteria for Reopening This Idea

Only take this seriously if:

• there is measurable performance pain
• profiling identifies real bottlenecks
• targeted optimization is not enough
• the rewrite cost is justified by roadmap needs

Current Recommendation

Do not pursue now.

Focus on:

• product scope
• discovery/indexing
• bundles/export/import
• routing evaluation / benchmarks
• launch / adoption

Revisit only when performance becomes a real blocker.