Weekly Analysis - 2026-04-18
Overview
The current board is small but already shows the intended operating pattern: one personal board spanning multiple repositories, with tasks that stay readable for humans and explicit enough for agents to route execution correctly.
Key Patterns
- Repository selection is now a first-class concern, not an implicit assumption.
- The best near-term wins are workflow infrastructure improvements rather than broad product work.
- Prompt reuse and visual consistency are both suffering from duplicated logic that wants to become shared structure.
Task Analysis
Polish task card project badges
tasksFrontendIn progress
Problem
Repository context is present in the task data model, but it still needs to feel obvious during scan-and-copy workflows.
Likely causes
- The board previously assumed a single project context.
- Project information competes with area and status instead of leading the card.
- The copy flow depends on the user noticing repository context early.
Relevant files/modules
- src/components/tasks/TaskCard.tsx
- src/components/tasks/TaskDetail.tsx
- src/components/home/HomePage.tsx
Suggested approach
- Render project as a dedicated badge on cards and detail pages.
- Keep the repo name exact and machine-readable.
- Make sure copied task prompts include `Project:` near the top.
Extract reusable color tokens for marketing pages
design-systemDesignTo do
Problem
Marketing pages are converging visually, but the palette still lives in scattered one-off values instead of reusable tokens.
Likely causes
- Fast page-building favored local overrides over shared primitives.
- There is no small approved palette slice for editorial/marketing surfaces.
- Similar accents and neutrals are being recreated page by page.
Relevant files/modules
- marketing theme tokens
- landing page section primitives
- shared CTA styles
Suggested approach
- Audit the repeated colors first, then collapse them into a minimal token set.
- Preserve expressive styling, but route it through stable semantic names.
- Validate the token pass on two or three existing pages before calling it finished.
Stabilize local skill loading across agents
openclaw-skillsBackendTo do
Problem
The same workflow prompt is still being pasted manually because local skill discovery is inconsistent across agents.
Likely causes
- Skill folder placement and metadata expectations differ between runtimes.
- Some agents support local skills while others only support globally installed ones.
- There is no single compatibility story documented for the workspace.
Relevant files/modules
- local skill folder conventions
- skill metadata readers
- agent bootstrap or discovery logic
Suggested approach
- Document the supported local-skill path first.
- Make taskboard skills portable with minimal metadata assumptions.
- Add one compatibility test path that proves the same skill can be reused across at least two agents.
Recommended Execution Order
- Finish the task-board repository-context improvements so every future handoff is clearer.
- Stabilize skill loading so prompt reuse becomes operational instead of aspirational.
- Extract design tokens once the board and agent workflow are less likely to churn.
Quick Wins
- Keep project names exact in every copied prompt.
- Add local skills next to the board so the prompt source of truth is versioned.
Risks
- If project names are prettified in the UI, agents may target the wrong repository.
- Skill portability may still vary across runtimes even after the board-side cleanup.