BuilderIO Skills
BuilderIO's skills repo packages agent-native review artifacts: visual plans before work, visual recaps after diffs, watchdog review, plan arbitration, and autonomy/budget utilities.
BuilderIO's skills repo is a small, composable skill library for coding agents. Its README highlights /visual-plan and /visual-recap as the default install choices; the repo also includes agent-watchdog, plan-arbiter, plow-ahead, efficient-frontier/fable, stay-within-limits, quick-recap, and read-the-damn-docs. Source: GitHub BuilderIO/skills, 2026-06-25
Current Source Snapshot
As of the 2026-07-06 registry refresh, BuilderIO/skills remains the source repo for the skill bodies, while the companion BuilderIO/agent-native package has latest @agent-native/skills@0.2.253. The earlier 2026-06-30 snapshot had BuilderIO/skills on main at 6c31f2188f87d4afdc38c14c3af8bed2088e4558, no releases/tags exposed by git ls-remote, MIT licensing, and GitHub reports of 2,911 stars and 151 forks. Recheck before pinning because Agent Native packages are moving quickly. Source: npm registry packument, 2026-07-06; git ls-remote, 2026-06-30; GitHub API, 2026-06-30
High-Signal Skills
| Skill | Use |
|---|---|
/visual-plan |
Turn an ordinary plan into a scannable artifact with diagrams, file maps, annotated code, open questions, and optional UI/prototype review. |
/visual-recap |
Turn a branch, commit, PR, or diff into an interactive recap with annotated diffs, API/schema summaries, file maps, and review notes. |
/agent-watchdog |
Audit another agent's work from a session, transcript, PR, branch, or run summary. |
/plan-arbiter |
Compare competing agent plans and pick one executable direction. |
/read-the-damn-docs |
Force authoritative docs checks before version-sensitive implementation. |
The visual-plan signal matters because it changes the artifact shape. Plans that are too important for chat should become inspectable, commentable artifacts with diagrams and UI flow storyboards. Steve Sewell's bookmarked demo framed /visual-plan as a way to see a full user flow as a storyboard instead of one screenshot at a time. Source: X/@Steve8708, 2026-06-25
The earlier demo artifact showed a run-history and plan page for "better plans for Claude Code," validating the same principle for Codex: planning should leave a visual, reviewable object with flow context and verification gates, not only a chat transcript. Source: X/@Steve8708, 2026-06-18
Visual Recap and Design-System Indexing
The July visual-recap bookmark adds a concrete GitHub PR artifact: "Add design system indexing from GitHub repos and npm packages." That moves /visual-recap from generic diff presentation toward reviewable design-system context. A useful recap should not only say which files changed; it should surface the UI contract, token/component provenance, package boundary, and any design-system drift created by the change. Source: X/@Steve8708, 2026-07-02
Local decision: keep visual recaps as Agent Legibility artifacts. Promote design-system indexing when the code change touches tokens, component APIs, stylesheets, npm package boundaries, or generated UI. Do not require it for ordinary backend-only diffs. Source: local artifact review, 2026-07-06
Local Adoption
Do not install the whole BuilderIO set by default. Kevin's local adoption is selective:
visual-plan is now an executable skill route for inspectable diagrams, flow maps,
storyboards, and verification gates before broad implementation.
Treat the rest of the BuilderIO set as candidate sources for two local routes:
- Agent Legibility and Proof-Carrying Work: visual plans/recaps make agent work reviewable before and after implementation.
- Agent Control Plane Architecture: watchdog, plan arbitration, usage-limit checks, and quick recaps are control-plane affordances for many-agent work.
Promotion gate for additional BuilderIO skills: audit the skill, choose hosted vs local-file artifact mode, add exact routing cases, and decide whether Kevin wants Agent-Native Plan as a dependency.
/visual-plan and /visual-recap should be treated as legibility tools, not another planning framework. Use them when the work would benefit from a durable review artifact: multi-screen user flows, file maps, diagrams, API/schema changes, or PR recaps. Ordinary one-file tasks should stay in chat and tests. Source: X/@Steve8708, 2026-06-25; Source: BuilderIO/skills README, 2026-06-30
Timeline
- 2026-07-06 | Added the visual-recap/design-system indexing bookmark. Local decision: use visual recaps when a diff needs UI/token/component provenance, and keep Agent Native package authority current at
@agent-native/skills@0.2.253. Source: X/@Steve8708, 2026-07-02; Source: npm registry, 2026-07-06 - 2026-07-01 | Refreshed Agent Native package snapshot:
@agent-native/skills@0.2.197, tag and HEAD83131cd52ae0c8c7857dd5665d80a901590780d1. Source: npm registry packument;git ls-remote, 2026-07-01 - 2026-06-30 | Deep-reviewed Steve Sewell's
/visual-planbookmark, inspected the local video thumbnail, resolvedBuilderIO/skillsandBuilderIO/agent-native, and recorded current source snapshots. Kept local adoption selective:visual-planis live; additional BuilderIO skills remain candidates behind audit and dependency review. Source: X/@Steve8708, 2026-06-25; Source: GitHubBuilderIO/skills; GitHubBuilderIO/agent-native, 2026-06-30 - 2026-06-29 | Promoted the local
visual-planexecutable skill route after the missing-skills audit. VerifiedBuilderIO/skillsHEAD at6c31f2188f87d4afdc38c14c3af8bed2088e4558; no tags exposed bygit ls-remote. Source: missing-skills audit, 2026-06-29 - 2026-06-30 | Added the earlier visual-plan demo tweet and local reviewed artifact as the concrete source for the Codex/Claude visual planning route. Source: X/@Steve8708, 2026-06-18
- 2026-06-25 | Page created after Kevin bookmarked Steve Sewell's
/visual-planuser-flow storyboard demo, with BuilderIO/skills and BuilderIO/agent-native as resolved artifacts. Source: X/@Steve8708, 2026-06-25; Source: GitHubBuilderIO/skillscommit735e6a29ea4e92ce33ae73cffa78a4cf3bd307fc