Skills Atlas
Skills are procedural memory. They preserve task triggers, prerequisites, steps, privacy rules, pitfalls, rollback, and verification independently of the current model. The count of 293 is a live inventory snapshot, not a permanent promise; skills are added, patched, archived, or removed as the system evolves.
Describe the exact request or condition that should load the skill.
Keep changing system state in canonical notes, not embedded forever in a skill.
List tools, credentials by reference, permissions, and source material.
Record stable pitfalls and recovery—not one-off task history.
Run the procedure and verify outputs before treating it as reusable.
24 verified skill categories
What breaks—and how to recover.
Consolidate overlapping narrow skills under a clear umbrella.
Patch the skill immediately after discovering a changed interface.
Publish purpose and safe examples, not endpoints, credentials, or sensitive operational steps.
Copyable implementation prompt
Write a portable skill for [TASK]. Include trigger, purpose, non-goals, prerequisites, inputs, numbered steps, privacy rules, pitfalls, rollback, verification, and references. Keep machine-specific values as placeholders.
Use placeholders for paths, accounts, endpoints, identities, and credentials. Never publish secret values, private records, or a live control surface.
Verification gate
- Trigger is precise
- Procedure is model-independent
- Privacy and rollback are explicit
- Commands or actions have been tested
- No temporary task state is embedded