microsoft/SkillOpt

View on GitHub

SkillOpt is a text-space optimizer that trains reusable natural-language skills for frozen LLM agents through trajectory-driven edits, validation-gated updates, and deployable best_skill.md artifacts.

Stars: 17.3k
Forks: 1.6k
Open beginner issues: 0
Indexed issues: 25
Avg merge: 2d 23h
Merged PRs (30d): 9

Dominant language: Python
License: MIT
Last GitHub push: Sep 5, 2026
Latest indexed: Sep 19, 2026
Contributing guide: Contributing guide
Code of conduct: Code of conduct
Beginner labels: help wanted

Beginner friendly Everything

Open All

25 open issues indexed

Release cut for the Aug adopt/webui hardening + 2 residual staging gaps Open

Difficulty 5/5 Over a week Newbie friendliness 42/100

microsoft/SkillOpt#288 · Opened Sep 20, 2026
skillopt-sleep Codex harvest ingests its own headless replay sessions Open

Difficulty 3/5 1-2 days Newbie friendliness 72/100

microsoft/SkillOpt#286 · 2 comments · Opened Sep 20, 2026
skillopt-sleep harvest crashes when a transcript disappears during discovery Open

Difficulty 2/5 1-3 hours Newbie friendliness 84/100

microsoft/SkillOpt#284 · 2 comments · Opened Sep 20, 2026
Proposal: optional offline typed-decision judge for rubric scoring (SemIf / NanoJev pattern) Open

Difficulty 5/5 Over a week Newbie friendliness 28/100

microsoft/SkillOpt#283 · Opened Sep 19, 2026
Tool suggestion: contract layer for skill optimization (open source) Open

Difficulty 5/5 Over a week Newbie friendliness 25/100

microsoft/SkillOpt#281 · Opened Sep 15, 2026
assign_splits's own non-empty-val guarantee doesn't fire when every real task hashes into test Open

Difficulty 3/5 1-2 days Newbie friendliness 65/100

microsoft/SkillOpt#271 · 2 comments · Opened Sep 6, 2026
skillopt-sleep: adopt() writes a backup that nothing can restore — add a `revert` command Open

Difficulty 4/5 3-5 days Newbie friendliness 68/100

microsoft/SkillOpt#247 · Opened Aug 22, 2026
[FYI] Automated, natural language Skillopt skill Open

Difficulty 5/5 Over a week Newbie friendliness 20/100

microsoft/SkillOpt#236 · Opened Aug 18, 2026
context length and trajectory problems Open

Difficulty 5/5 Over a week Newbie friendliness 30/100

microsoft/SkillOpt#233 · Opened Aug 17, 2026
Proposal: add a version-gated DeepSeek Harness transcript source to SkillOpt-Sleep Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

microsoft/SkillOpt#225 · Opened Aug 14, 2026
section_present rejects numbered or annotated headings, and the optimizer 'fixes' it by forbidding the skill's own heading format Open

Difficulty 4/5 3-5 days Newbie friendliness 48/100

microsoft/SkillOpt#175 · 2 comments · Opened Jul 26, 2026
Several reported scores are incompatible with the released test split sizes Open

Difficulty 3/5 1-2 days Newbie friendliness 48/100

microsoft/SkillOpt#158 · 2 comments · Opened Jul 19, 2026
LiveMath scores imply 125 test items, but the released split has 124 Open

Difficulty 3/5 1-2 days Newbie friendliness 45/100

microsoft/SkillOpt#157 · 1 comment · Opened Jul 19, 2026
Proposal: community rule library with local empirical gating (privacy-preserving federated skill transfer) Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

microsoft/SkillOpt#156 · 1 comment · Opened Jul 19, 2026
Proposal: AgenticReplayBackend — replay mined tasks with a real agentic CLI in a throwaway worktree, gate on the diff Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

microsoft/SkillOpt#155 · 5 comments · Opened Jul 19, 2026
Mining is limited to programmatic success checks — intent-level improvements get dropped or Goodharted into shallow proxy rules Open

Difficulty 5/5 Over a week Newbie friendliness 30/100

microsoft/SkillOpt#154 · 3 comments · Opened Jul 19, 2026
Help wanted: integrate SkillOpt with Superpowers for measurable agent efficiency Open

enhancement help wanted

Difficulty 4/5 3-5 days Newbie friendliness 48/100

microsoft/SkillOpt#132 · 9 comments · Opened Jul 13, 2026
[Request] Release optimized skill artifacts for additional methods / models / harnesses Open

Difficulty 5/5 Over a week Newbie friendliness 35/100

microsoft/SkillOpt#130 · Opened Jul 13, 2026
SkillOpt-Sleep: target existing hand-written skills, not just the dedicated managed skill Open

Difficulty 5/5 Over a week Newbie friendliness 38/100

microsoft/SkillOpt#120 · 1 comment · Opened Jul 10, 2026
A Confusion Regarding s_cur and s_best Open

Difficulty 2/5 1-3 hours Newbie friendliness 45/100

microsoft/SkillOpt#113 · 1 comment · Opened Jul 9, 2026
Explore SkillOpt for improving Cursor agent skills from our session data Open

Difficulty 5/5 Over a week Newbie friendliness 28/100

microsoft/SkillOpt#111 · 2 comments · Opened Jul 8, 2026
How is the baseline evaluation score aggregated across runs? (single-seed vs multi-seed / best-of-N) Open

Difficulty 3/5 1-2 days Newbie friendliness 45/100

microsoft/SkillOpt#108 · 2 comments · Opened Jul 7, 2026
SkillOpt-Sleep: Support harvest from Vscode/ Visual studio copilot sessions Open

Difficulty 4/5 3-5 days Newbie friendliness 35/100

microsoft/SkillOpt#97 · 3 comments · Opened Jul 2, 2026
question about train config Open

Difficulty 2/5 1-3 hours Newbie friendliness 50/100

microsoft/SkillOpt#75 · 1 comment · Opened Jun 22, 2026
Release optimized skills and benchmark datasets on Hugging Face Open

Difficulty 4/5 3-5 days Newbie friendliness 52/100

microsoft/SkillOpt#10 · 1 comment · Opened May 26, 2026

microsoft/SkillOpt

Get new issues in your inbox