Phase F: skill-eval description optimizer + HTML viewer (+ telegram isolation, applied map-plan description)#159

Merged

azalio merged 11 commits into

mainfrom

Jun 4, 2026

Commits on Jun 4, 2026

Phase F ST-001: OptimizeIterationRecord + OptimizeResult + ProposerFn in eval_schema

azalio
and
claude
committed
Phase F ST-002: proposer.py — claude -p subprocess description proposer

azalio
and
claude
committed
Phase F ST-003: description_optimizer.py — anti-overfit held-out selection

azalio
and
claude
committed
Phase F ST-004: viewer.py — jinja2 HTML report from OptimizeResult

azalio
and
claude
committed
Phase F ST-006: apply_patcher.py — frontmatter --apply patcher + render orchestration

azalio
and
claude
committed
Phase F ST-005: optimize + view CLI subcommands on skill-eval

azalio
and
claude
committed
Phase F ST-007: 3 optimizer eval-set fixtures (>=8 entries) + authoring note

azalio
and
claude
committed
Phase F ST-008: document optimize/view subcommands in map-skill-eval docs

azalio
and
claude
committed
Phase F ST-009: extend no-anthropic guard to all 4 modules; full gate green

azalio
and
claude
committed
fix(skills-eval): isolate eval claude -p from the telegram-bridge listener

azalio
and
claude
committed
Phase F: apply optimized map-plan description + correct the description cap to the spec (1024)

azalio
and
claude
committed