Skip to content

Team workspace improvement choices

Implemented in this pass

  • Illustrated basecamp, desktop navigation rail and mobile navigation; personal progress remains above today's tasks. No invented achievements or demo tasks.
  • Docs pull-request CI for workflow tests, record validation, scripts, strict build and repository checks. No secrets, write token or deployment step.
  • Day-one readiness: engineer prerequisites, concrete backend gaps, Niwesh's test environment, and a first-week rehearsal.
  • Task-relevant learning guidance, optional Friday retrieval questions and a short retrospective. Editor choice stays flexible.

Keep as proposals

  • Delegated acceptance and parallel tasks during review: these change agreed authority and work-in-progress rules. Discuss with the founder using actual review waiting evidence first.
  • Auto-merge and live branch aggregation: records can contain authoritative decisions. Unmerged branch data also needs conflict handling and provenance. Do not present it as accepted team progress.
  • Chat bots and paid deployment changes: need a selected service, account access and explicit publishing/spending authorization.
  • Team hike reward: a good optional real-world celebration; founder chooses budget and timing. No promised reward has been recorded.
  • Automatic hours: session timestamps include breaks and waiting. At most suggest a draft duration and ask the person to confirm; never record it as fact.
  • Read-only dashboard: keep current editing until the team tests the terminal workflow. Removing the fallback now would reduce available ways to report work.
  • Screenshot compression: useful future work, but it must preserve legibility and attachment hashes. Do not rewrite existing evidence during a visual pass.
  • Native Claude shortcuts: AGENTS.md already provides the portable fallback. Verify discovery in a real Claude session before adding tool-specific symlinks.

Review after the first week

Ask each member what slowed them down, which instruction was unclear, and which AI answer they had to correct. Pick one improvement for the next week. Members may propose skill edits by pull request; normal review and founder merge apply. Measure actual waiting stages before changing the one-task rule.

Verification for this pass

  • Workflow tests: 67 passed; record validation, strict MkDocs build and repository hooks passed locally.
  • Browser: Chromium desktop at 1440 × 1000; phone at 390 × 844; narrow phone at 320 × 780.
  • Today and Academy had no horizontal overflow at the checked phone widths.
  • At 390 px, the task card starts around 635 px from the top, after personal progress.
  • Engineer navigation hides the manager view; Niwesh can open it. This is UI routing, not authorization.
  • Existing records were read without adding fictional progress. Connected-folder writes, the product's backend/frontend startup and real teammate AI logins were not exercised in this pass.
  • GitHub-hosted CI remains unverified until this workflow is pushed and runs.