Standup

Generated with timbers draft standup --since 2026-02-13 --until 2026-02-13 | claude -p --model opus Fixed a silent bug in the PostToolUse hook — it read $TOOL_INPUT (always empty) instead of stdin, meaning the post-commit reminder was a no-op since it was created Three CI fixes stacked up: stale git-notes fetch referencing storage timbers no longer uses, Hugo baseURL pointing at old org, and GITHUB_TOKEN pushes not triggering downstream deploys. Chained devblog→pages deploy to close the loop. Two pre-existing test failures unblocked: TestRepoRoot broke in worktree environments, TestCommitFiles used live HEAD which returns empty on merge commits. Both now own their test state. Released v0.8.0 and v0.9.0 — manual release flow required because claude -p can’t nest inside Claude Code sessions Rewrote coaching system informed by Opus 4.6 prompt guide: motivated rules (WHY behind each rule), concrete 5-point notes trigger checklist, XML tag structure. Council confirmed no model-specific variants needed — clear coaching IS Opus-optimized coaching. Added AgentEnv interface with registry pattern so future agent environments (beyond Claude) are self-contained via init() registration. Refactored init, doctor, setup, uninstall to use it. Shipped marketing landing page for the Hugo site — competitive positioning matters with Entire.io in the landscape. Follow-up polish pass on terminal blocks. --notes flag was shipped in v0.8.0 but undocumented everywhere except prime coaching — backfilled across README, tutorial, spec, and agent-reference Regenerated example artifacts using 55 real entries instead of 30 backfilled ones — decision-log and sprint-report output quality jumped noticeably with richer why and notes data