agentleFS
BuilderIOBuilderIO/agent-native/.agents/skills/visual-answer/SKILL.md

visual-answer

Skill7.6k starsChanged 5 months agoMITView on GitHub

Answer code/product questions as visual Plan artifacts using repo, bridge, or GitHub context; use for API specs, UI look, schema models, and architecture.

We split this skill into the rules it gives an agent

9 rules. About 0 are common, 3 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill when the user asks:

    line 20
  2. Very specific:

    Call get-plan-blocks before authoring.

    line 40
  3. Very specific:

    If the question asks what components are available, call list-plan-components.

    line 41
  4. Very specific:

    When running from a terminal or coding-agent shell, write visual-answer-source.json:

    line 61
  5. Very specific:

    Use --source-url for a GitHub file/PR/commit/issue URL, --prev-plan-id to refresh an existing answer, and --visibility private for …

    line 83
  6. Very specific:

    Do not use visual-answer for a plain implementation plan.

    line 89
  7. Very specific:

    Do not use it for a PR diff recap.

    line 91
  8. Distinctive:

    Do not skip code inspection and infer from names alone.

    line 92
  9. Distinctive:

    Do not publish screenshots, secrets, credential-looking values, or private source excerpts beyond what the user has asked to visualize.

    line 93

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Visual Answer

/visual-answer turns a specific code or product question into a published Agent-Native Plan artifact. It is for questions that need a visual, inspectable answer rather than a chat paragraph: API contracts, schema/data models, UI states, component behavior, architecture flows, and code evidence.

When To Use

Use this skill when the user asks:

  • "what is the API spec for this?"
  • "what does this UI look like?"
  • "what is the schema model for x?"
  • "draw the flow for this code path"
  • "show me the current shape of this component/API/data model"

For history questions about what changed or shipped, search merged PR recaps first with search-pr-recaps, then pull a relevant hit into the conversation with show-visual-plan (it renders that plan or recap's blocks inline). Reach for get-visual-plan only when you need the full bundle/MDX to edit, not to display. Use visual-answer when the current codebase, a local bridge, or GitHub source needs to be inspected to produce a new answer.

Workflow

  1. Inspect the real source first. Use the host agent's repo tools, the Plan local bridge, or GitHub/source links. Do not invent endpoints, schema fields, UI states, file names, or behavior.
  2. Call get-plan-blocks before authoring. Use the live registry, not memory. If the question asks what components are available, call list-plan-components.
  3. Pick the evidence blocks:
    • API shape: openapi-spec plus api-endpoint; use before/after only when comparing historical changes from recaps.
    • UI look: wireframe or a columns before/after pair when comparing a recap.
    • Schema/data model: data-model, optionally with diagram.
    • Code evidence: file-tree, tabs, annotated-code, and diff when the answer depends on implementation details.
  4. Publish with the Plan visual-answer action. Include the user's question, repoPath/sourceUrl when known, a concise title/brief, and MDX source under mdx.
  5. In Agent-Native chat the published answer renders inline automatically — its blocks appear in the conversation. Add a one-line summary of the evidence plus the deep link; do not paste raw MDX or block source as text. (In terminal or external MCP hosts that cannot render the blocks, return the URL.)

Terminal Handoff

When running from a terminal or coding-agent shell, write visual-answer-source.json:

{
  "question": "What is the billing API shape?",
  "title": "Billing API visual answer",
  "brief": "Shows the request and response contract.",
  "repoPath": "owner/repo",
  "sourceUrl": "https://github.com/owner/repo",
  "mdx": {
    "plan.mdx": "---\ntitle: Billing API visual answer\n---\n\n..."
  }
}

Then publish:

agent-native plan visual-answer publish --question "What is the billing API shape?" --source visual-answer-source.json --repo owner/repo

Use --source-url for a GitHub file/PR/commit/issue URL, --prev-plan-id to refresh an existing answer, and --visibility private for owner-only output. The command writes visual-answer-url.txt.

Don't

  • Do not use visual-answer for a plain implementation plan; use visual-plan.
  • Do not use it for a PR diff recap; use visual-recap.
  • Do not skip code inspection and infer from names alone.
  • Do not publish screenshots, secrets, credential-looking values, or private source excerpts beyond what the user has asked to visualize.

Related Skills

  • visual-plan
  • visual-recap
  • delegate-to-agent
  • context-awareness

Download now (ZIP)

Trending

Steady: 0 new projects took it in the last 4 weeks, none the 4 weeks before.thin data

Week of Aug 30: 0 new projectsWeek of Sep 6: 0 new projectsWeek of Sep 13: 0 new projectsWeek of Sep 20: 0 new projectsWeek of Sep 27: 0 new projectsWeek of Oct 4: 0 new projectsWeek of Oct 11: 0 new projectsWeek of Oct 18: 0 new projects
Aug 2026Sep 2026Oct 2026
New projects taking a version of this file each week, over the 8 weeks our latest run read: only 0 in all, so this is a hint, not a trend.

Who's talking about BuilderIO/agent-native

The biggest 5 of 61 mentions linking to the repository that our crawl found. Each line is their own headline.

See all 61

Talked about on GitHub and Bluesky (61 times).

On GitHub · 50

45 more on GitHub

On Bluesky · 11

6 more on Bluesky

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.