visualize
Create source-adjacent HTML visualizations for /visualize requests, docs, markdown, standardized plans, process flows, charts, Mermaid diagrams, context maps, and recommendation diagrams.
How do I install this agent skill?
npx skills add https://github.com/buiducnhat/cobrew --skill visualizeIs this agent skill safe to install?
- Gen Agent Trust Hubpass
The skill is safe to use. It creates HTML visualizations from project documentation and markdown files. It correctly includes security best practices such as escaping source text to prevent injection in the generated HTML. It uses a well-known CDN to load Mermaid diagrams and uses a standard shell command for timestamps. A low-risk surface for indirect prompt injection exists because it processes user-provided content.
- Socketpass
No alerts
- Snykwarn
Risk: MEDIUM · 1 issue
What does this agent skill do?
Visualize
Overview
Create source-adjacent HTML visualizations from repository documentation, markdown files, standardized plan folders, and arbitrary user context. This skill is Huashu-derived in workflow discipline, but it is intentionally documentation-focused: it turns real source material into readable maps, timelines, flows, risk blocks, and recommendation diagrams.
Scope
Use this skill for:
- Documentation and markdown visualization
- Standardized plan folder visualization
- Flow, chart, Mermaid, context map, and recommendation diagram requests
- Source-adjacent HTML summaries that help users inspect structure, risks, decisions, and next steps
Do not use this skill for:
- Brand sites or marketing pages
- Product prototypes or production web apps
- Slide decks, long motion demos, or broad visual design systems
Output Conventions
- Plan folder: create
visualize.htmlandvisualize-assets/inside the plan folder. - Brainstorm folder: create
visualize.htmlandvisualize-assets/inside the brainstorm folder. - Current context tied to an active plan or brainstorm artifact folder: create
visualize.htmlandvisualize-assets/inside that artifact folder. - Markdown or document file: create
<source-base>.visualize.htmland<source-base>.visualize-assets/beside the source file. - Source-less context with no associated artifact folder: create
docs/.visualizations/<slug>-YYMMDD-HHmm/visualize.htmlanddocs/.visualizations/<slug>-YYMMDD-HHmm/visualize-assets/unless the user specifies a target. Generate the timestamp withdate +%y%m%d-%H%M.
Always copy references/templates/visualize-theme.css into the adjacent assets folder and link the copied local CSS from the generated HTML. Mermaid CDN usage is allowed for diagrams.
Workflow
- Load project context when visualizing repository files (if the current session does not already have it).
- If
docs/SUMMARY.mdexists, read it first. - Load only task-relevant detail docs.
- Prioritize
Code Standarddocs for implementation conventions. - If docs conflict with code or user intent, use the available input/question tool before broad changes.
- If
- Load the source material.
- For plan folders, read
SUMMARY.mdand relevantphase-XX-*.mdfiles. - For brainstorm folders, read
SUMMARY.mdand relevantsection-XX-*.mdfiles. - For markdown or document files, read the requested source directly.
- For current context tied to an active plan or brainstorm artifact folder, use the conversation/request content plus the artifact folder as the output target.
- For source-less user context with no associated artifact folder, use only the provided conversation/request content.
- For plan folders, read
- Identify the visualization intent.
- Determine whether the user needs a plan map, document map, process flow, decision view, recommendation view, or mixed visualization.
- Ask with the input/question tool only when the source, target, or output ambiguity cannot be resolved safely.
- Build a compact source inventory before choosing layout.
- Use
references/workflow.mdandreferences/content-patterns.md. - Capture the source hierarchy, entities or files, phases or tasks, dependencies, decisions, risks, blockers, recommendations, verification steps, and missing evidence when present.
- Treat templates as shells. A block belongs in the output only when it summarizes source-backed content, clarifies a relationship, or exposes a meaningful gap.
- Use
- Choose the template with
references/router.md.- Plan folders use
references/templates/plan.html. - Brainstorm folders use
references/templates/context.html. - Single documents use
references/templates/document.html. - Source-less context uses
references/templates/context.html.
- Plan folders use
- Extract facts and assumptions into the selected block set.
- Preserve source facts, file paths, statuses, risks, decisions, verification steps, and recommendations.
- Label assumptions clearly when the source leaves gaps.
- Compose the HTML.
- Start from the selected template.
- Replace
VISUALIZE:marker regions with source-backed content and remove blocks that do not earn their place. - Escape source-derived text before inserting it into HTML unless the text is intentionally authored generated markup.
- Copy
visualize-theme.cssbeside the output before linking it.
- Add Mermaid diagrams where they clarify structure.
- Use
references/mermaid-recipes.md. - Prefer several small diagrams over one dense diagram.
- Skip Mermaid when a diagram would only restate headings or duplicate nearby prose.
- Use readable HTML/CSS fallback blocks when Mermaid syntax is uncertain.
- Use
- Verify the output with
references/verification.md.
Rules
- Do not invent facts, fake metrics, fake decisions, fake users, fake constraints, or unsupported recommendations.
- Do not hide missing source content. Add an assumption or warning block instead.
- Keep visualizations factual, source-adjacent, and easy to inspect.
- Keep templates and generated HTML focused on documentation visualization, not broad design/prototype work.
- Avoid large inline style blocks; use the copied fixed theme CSS.
- Escape source-derived text before writing it into HTML unless it is intentional generated markup authored by the agent.
- Preserve source links and paths when they help the user trace content back to origin.
How can the creator link this skill?
Add the canonical catalog link to the repository README so users can inspect current installs and available audits. The publishing guide covers the complete discovery path.
<a href="https://skillzs.dev/skills/buiducnhat/cobrew/visualize">View visualize on skillZs</a>