| name | devcard |
| description | Orchestrate DevCard MCP tools for GitHub profile analysis. Use when user wants to
check a developer profile, compare developers, audit GitHub presence, generate a
DevCard, or improve their profile. Entry point for all DevCard MCP interactions.
Not for CLI-only operations — use devcard-generate skill for that.
|
DevCard — MCP Tool Orchestration
DevCard's MCP server exposes 10 tools for analyzing, scoring, and improving GitHub
profiles. This skill ensures you use them in the right order and handle edge cases.
When to Use
- User asks to "check", "analyze", or "look at" a GitHub profile
- User wants to compare two developers
- User mentions DevCard, profile scores, or agent readiness
- User wants to fix profile issues or improve their GitHub presence
- NOT when user just wants to generate a card file — use
/devcard-generate
- NOT when user only needs agent-readiness — use
/devcard-agent-ready
Available Tools
Read-Only (no token needed)
| Tool | Purpose | Returns |
|---|
get_devcard | Full profile as structured JSON | DevCard with 14+ signals |
get_developer_summary | Concise agent-friendly text | One-paragraph summary |
compare_developers | Side-by-side with scores | Comparison with quality deltas |
check_developer_stack | Technology lookup | Found/not-found per tech |
render_card | Visual SVG card | SVG string (5 themes) |
Analysis (no token needed)
| Tool | Purpose | Returns |
|---|
audit_profile | Score + find issues | human_score, agent_score, issues[] |
analyze_repo | Deep-dive single repo | classification, issues, suggestions |
Write (token required, dry_run by default)
| Tool | Purpose | Changes |
|---|
fix_profile | Fix profile-level issues | README, devcard.json, descriptions |
fix_repo | Fix repo-level issues | AGENTS.md, topics, description |
agent_ready | Full audit + fix cycle | All of the above |
Process
- Start with
audit_profile to get the dual scores and issue list.
- Present scores honestly. Human visibility (0-100) and agent readiness (0-100). Don't sugarcoat.
- Show top 3-5 issues sorted by severity (high → medium → low).
- Ask the user which issues they want to address. Don't auto-fix.
- Preview with
dry_run=true before any write operation. Show exactly what will change.
- Apply only after explicit confirmation with
dry_run=false.
- Re-audit to show score improvement after fixes.
Common Rationalizations
| Shortcut | Why It Fails |
|---|
| "Skip audit, just fix everything" | User needs to see what's wrong first. Blind fixes erode trust. |
| "Apply fixes without preview" | Dry-run exists for a reason. One bad description update is hard to revert. |
| "Combine multiple fix types in one call" | Each fix type should be previewed separately so user can accept/reject individually. |
Anti-patterns
- DO NOT call
fix_profile or fix_repo without showing dry_run=true preview first
- DO NOT skip
audit_profile and jump straight to fixes
- DO NOT inflate scores or hide bad signals — the user needs honest data
- DO NOT use write tools without confirming the user has a GitHub token set