Anti-Patterns

1. Probe-first auth by remote call

Avoid adding extra remote probe calls just to test auth readiness. Use local binding checks first, then validate through the first real operation call.

2. Dynamic link renaming at runtime

Do not teach runtime logic to append suffixes or auto-rename command names. Link naming is a skill author decision at design time.

3. Overweight wrapper skill

Do not duplicate generic OAuth/error docs in each provider skill. Keep wrappers thin and reference skills/uxc for shared behavior.

4. Ambiguous invocation style

Avoid mixed, obsolete, or conflicting argument styles in examples. Prefer key=value, and include bare JSON positional only where it adds clarity.

5. Write-before-read flows

Do not default to write operations. Read current state first, then require explicit user confirmation before destructive or high-impact writes.

6. Host-only assumptions

Do not infer protocol/path/auth only from host shape. Always verify with official docs search and uxc probing before publishing wrapper commands.