SkillOpt · Issues· 42 open

Proposal: AgenticReplayBackend — replay mined tasks with a real agentic CLI in a throwaway worktree, gate on the diff

Tool suggestion: contract layer for skill optimization (open source)

assign_splits's own non-empty-val guarantee doesn't fire when every real task hashes into test

[FYI] Automated, natural language Skillopt skill

Mining is limited to programmatic success checks — intent-level improvements get dropped or Goodharted into shallow proxy rules

Help wanted: integrate SkillOpt with Superpowers for measurable agent efficiency

skillopt-sleep: adopt() writes a backup that nothing can restore — add a `revert` command

context length and trajectory problems

How is the baseline evaluation score aggregated across runs? (single-seed vs multi-seed / best-of-N)

Proposal: add a version-gated DeepSeek Harness transcript source to SkillOpt-Sleep

section_present rejects numbered or annotated headings, and the optimizer 'fixes' it by forbidding the skill's own heading format

Several reported scores are incompatible with the released test split sizes

LiveMath scores imply 125 test items, but the released split has 124

Proposal: community rule library with local empirical gating (privacy-preserving federated skill transfer)

A Confusion Regarding s_cur and s_best