| name | devcard-agent-ready |
| description | Make a GitHub profile readable by AI agents and coding assistants. Use when user
wants to improve their "agent readiness score", deploy devcard.json, create llms.txt,
generate AGENTS.md, or make their profile machine-readable for tools like Claude Code,
Cursor, Copilot, or Windsurf. Covers the full agent-readiness optimization workflow.
|
DevCard Agent-Ready
AI coding assistants (Claude Code, Cursor, Copilot, Windsurf) increasingly read
developer profiles to understand who they're working with. This skill optimizes
a GitHub profile for machine readability — structured data, agent instructions,
and LLM-friendly documentation.
When to Use
- User asks about "agent readiness" or "agent score"
- User wants to deploy devcard.json, llms.txt, or AGENTS.md
- User asks "how do AI tools see my profile"
- User wants to improve their profile for coding assistants
- NOT for human-facing profile improvement — use
/devcard-profile-audit
- NOT for generating visual cards — use
/devcard-generate
Prerequisites
- DevCard installed:
uv sync in the devcard project
- GitHub token with repo write access for deploying files
- The user's profile repo must exist (github.com/username/username)
What Makes a Profile Agent-Ready
| Signal | Points | What It Is |
|---|
devcard.json in profile repo | 20 | Structured identity data for agents |
AGENTS.md in top repos | 20 | Instructions for coding agents |
llms.txt in profile repo | 10 | LLM-friendly profile summary |
| Structured READMEs | 15 | Parseable documentation |
| Dependency files | 10 | Stack detection (pyproject.toml, package.json) |
| Topics on repos | 10 | Domain classification signals |
| Classification coverage | 10 | Repos tagged as library/tool/app |
| CI/CD adoption | 5 | Build automation signals |
Process
-
Check current agent readiness score.
uv run devcard advise <username> --format json --no-cache
Look at agent_score in the output. Or via MCP: audit_profile(username).
-
Identify the biggest gaps. The score breakdown shows which signals are missing.
Prioritize by points: devcard.json (20) > AGENTS.md (20) > llms.txt (10).
-
Deploy devcard.json. This is the highest-value single action.
fix_profile(username, fixes=["devcard_json"], dry_run=true)
fix_profile(username, fixes=["devcard_json"], dry_run=false)
-
Generate AGENTS.md for top repos.
fix_repo(owner, repo, fixes=["agents_md"], dry_run=true)
fix_repo(owner, repo, fixes=["agents_md"], dry_run=false)
-
Deploy llms.txt.
uv run devcard generate <username> --format llms-txt > llms.txt
-
Add topics to repos that are missing them.
fix_repo(owner, repo, fixes=["topics"], dry_run=true)
-
Re-check score. Run the audit again and compare.
Common Rationalizations
| Shortcut | Why It Fails |
|---|
| "devcard.json is just metadata, not important" | It's 20% of the agent score. Agents use it to understand you instantly. |
| "AGENTS.md is overkill for my repos" | Coding agents read AGENTS.md before touching your code. Without it, they guess. |
| "I'll add topics later" | Topics are how agents classify your repos. No topics = invisible to automated discovery. |
| "My README is good enough for agents" | Agents parse structure (headings, code blocks), not prose. Structured > well-written. |
Anti-patterns
- DO NOT deploy devcard.json without previewing the content first
- DO NOT skip AGENTS.md for repos that coding agents might work on
- DO NOT generate llms.txt with stale data — always use
--no-cache
- DO NOT assume the profile repo exists — check first, create if needed
Verification Checklist