v0.1.0

v0.1.0 — First feature-complete release

Core

• PERT three-point estimation engine with task classification (XS–XL)
• Modifier stack: spec clarity, warm context, agent fit
• METR per-model p80 reliability threshold warnings
• Wave planner with DAG validation, topological sort, and LPT bin packing
• Self-calibration via SQLite observation store

CLI

• estimate — single task, file input (--file), or GitHub issues (--issues)
• calibrate — compare estimates against actuals
• validate — check observation YAML files
• Modifier flags: --spec-clarity, --warm-context, --agent-fit
• Auto warm-context detection from dispatch history (--history-file)
• Output formats: markdown (default) and JSON (--format json)
• Review mode selection: --review-mode none|self|2x-lgtm

Integrations

• GitHub adapter via gh CLI for issue fetching
• Claude Code plugin structure (.claude-plugin/plugin.json)
• /estimate skill wrapper for Claude Code

Quality

• 323 tests (unit + integration), CI on Python 3.10–3.12
• Ruff linting
• PyPI-ready packaging via hatchling