Sudhanva Agent Skills
This repository contains three small, read-only Agent Skills.
Use them to research the public work of Sudhanva Narayana. Each skill reads canonical
sudhanva.me pages and APIs. Each skill also tells the agent not to invent unpublished
details.
Install
Install all three skills:
npx skills add nsudhanva/sudhanva-agent-skills --all
To install only one skill, use this command. The CLI asks you to select the skill:
npx skills add nsudhanva/sudhanva-agent-skills
Included skills
| Skill | Purpose |
|---|
research-sudhanva-profile | Research the public professional profile and experience. |
find-production-ml-case-studies | Find and compare published production ML case studies. |
find-technical-writing | Search and cite published technical writing. |
Layout
The repository uses the Agent Plugins package
layout. Thus, plugin hosts and the skills CLI find the skills in skills/:
sudhanva-agent-skills/
βββ plugin.json
βββ mcp.json
βββ skills/
βββ find-production-ml-case-studies/SKILL.md
βββ find-technical-writing/SKILL.md
βββ research-sudhanva-profile/SKILL.md
Source and safety
- All referenced data is public and read-only.
- Use canonical pages and API responses in preference to search snippets.
- Keep the qualifiers on quantitative outcomes.
- Do not infer private contact details, availability, compensation, client
names, or unpublished infrastructure.
- The canonical discovery index is
/.well-known/agent-skills/index.json.
plugin.json and mcp.json put these skills
and the public-data and profile-insight MCP servers into one
Agent Plugin.
License
MIT. See LICENSE.
Validation
Run the same checks that CI uses. These checks have no dependencies:
The test suite checks the plugin and MCP manifests and the canonical project URLs. For each
skill, it checks the frontmatter, the directory name, the top-level heading, and the README
entry. The suite also fails if a SKILL.md file is in a location other than
skills/<name>/SKILL.md.