Skip to content
Permalink

Comparing changes

Choose two branches to see what’s changed or to start a new pull request. If you need to, you can also or learn more about diff comparisons.

Open a pull request

Create a new pull request by comparing changes across two branches. If you need to, you can also . Learn more about diff comparisons here.
base repository: microsoft/SkillOpt
Failed to load repositories. Confirm that selected base ref is valid, then try again.
Loading
base: main
Choose a base ref
...
head repository: mithudso/SkillOpt
Failed to load repositories. Confirm that selected head ref is valid, then try again.
Loading
compare: main
Choose a head ref
Checking mergeability… Don’t worry, you can still create the pull request.
  • 3 commits
  • 34 files changed
  • 2 contributors

Commits on Oct 5, 2026

  1. Bootstrap fork docs, add llms usage files and generic skilltask env 

    - repo-bootstrapper pass (additive): CLAUDE.md, AGENTS.md, GEMINI.md,
  copilot-instructions, memory/prompts logs, docs/ architecture suite,
  codebase overview, high-signal file index, audit report.
- llms.txt family documenting install, train/eval, config, backends,
  SkillOpt-Sleep, custom envs, outputs; scripts/build_llms_full.py.
- skillopt/envs/skilltask + configs/skilltask + scripts/skillopt_skill.py:
  train any Markdown skill against a JSONL task set graded by match or
  LLM judge, without editing upstream train.py/eval_only.py.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01QFucAGAd382ZruyUWH2B33
    @mithudso @claude
    mithudso and claude committed Oct 5, 2026
    Configuration menu
    Copy the full SHA
    a6382fb View commit details
    Browse the repository at this point in the history

  2. skilltask: prepend no-tools answer preamble to task inputs 

    claude_chat runs the target inside a claude -p agent harness with tools
denied. On a 38-task skill-optimizer run, several rollouts answered
'Bash and Read were denied' instead of the task, and reflection learned
a 'blocked-run reply' patch from that noise. env.answer_preamble
(default: built-in note, "" to disable) tells the target to answer in
text.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01QFucAGAd382ZruyUWH2B33
    @mithudso @claude
    mithudso and claude committed Oct 5, 2026
    Configuration menu
    Copy the full SHA
    f4232fe View commit details
    Browse the repository at this point in the history

  3. docs(fork): bind Stele project, fix evalkit subcommand, log v0.2-v0.3 

    - Add .stele/project.json binding the repo to Stele project skillopt-ocj5r
  (backfilled from docs, code, git history and upstream issues/PRs).
- CLAUDE.md and docs/COMPONENTS.md named the Sleep A/B subcommand `eval`;
  the CLI registers it as `evalkit` (skillopt_sleep/__main__.py:885).
- memory.md/prompts.md: retroactive v0.2 entry for f4232fe (skilltask
  no-tools answer preamble) and v0.3 for the Stele backfill.

Co-Authored-By: Claude Opus 5.5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01QtB7JJLPmYfH3W4tetDh9M
    @mithudso @claude
    mithudso and claude committed Oct 5, 2026
    Configuration menu
    Copy the full SHA
    86739ad View commit details
    Browse the repository at this point in the history
Loading