Turn an ad-hoc one-off action into a permanent skill. Then nudge the trigger so the skill fires on creation, not just review.
Takeaway 1
Any one-off interaction can become a permanent skill — "save that rule as a skill" is enough.
Takeaway 2
Skills capture not just the rule but when it fires — the trigger scope is its own design knob.
Takeaway 3
The skill list itself can be exported to a Markdown reference — perfect handoff documentation.
Save an Action as a Skill
The shortcut is: do the work once in chat, then ask the agent to save the result as a skill. In the demo the author asks the agent to audit a folder's file names. It runs the audit and reports back. The author then says: "Save that rule as a skill for all new files as well as to review existing files."
Save-as-skill turns prototyping into a deliverable. You explore an idea in chat; once it works, one prompt promotes it to a reusable, named skill — like saving a macro in older productivity tools.
The agent extracts the rule from the conversation, picks a name (the demo's became "file naming audit"), and writes the skill. From then on, any matching utterance fires the saved rule instead of re-deriving it.
"I need to re-describe the skill from scratch when I save it." You don't. The conversation that produced the work is the skill specification. The agent inherits everything it just learned about your intent.
Naming Schemas — A Rule Worth Saving
A naming schema is a structured rule for file names — for example, Month-Year-Team-Purpose. The author defines one in chat, audits an existing folder against it, then saves the audit logic as the file naming audit skill so every future file inherits the check.
Schemas are great save-as-skill candidates because the rule is precise and the work repeats. Define it once, audit forever, with no per-file thinking required.
The skill can be edited later — say, swap out a delimiter, or add a "Confidentiality" segment — by asking the agent to update the rule. Existing files won't be renamed, but the audit will start flagging anything that doesn't match the new schema.
Modifying the Trigger Scope
By default the file naming audit skill ran after files existed. The author asks: "Edit the file naming audit skill to make sure it also applies when creating new files." That changes when the skill fires — not the rule itself, but the trigger scope.
Trigger scope is independent of the skill's content. You can broaden, narrow, or relocate the moment a skill activates without touching the rule it enforces. It's the front-loading move that turns audit into prevention.
After the edit, when the author later asks the agent to create a Markdown readme in the document library, the skill kicks in proactively — proposing a name that matches the schema before the file is saved. The user can accept it or refine.
"Editing the trigger means rewriting the skill." It doesn't. The trigger is the "when," the body is the "what." You can change either independently. Front-loading triggers is one of the cheapest wins in skill design.
Generating a Skill Reference
"Can you create a Markdown readme for all the skills you have installed and save it right here in this document library?" The author runs that prompt and the agent produces a "March 2026 admin skills reference" — every skill, with its purpose and when it fires, formatted as Markdown.
The agent treats the skill catalogue as data. You can summarize it, export it, audit it, hand it off — all from chat. No admin export tool, no manual write-up.
This is also how the file naming audit skill demonstrated its trigger scope at work — the new file (the readme itself) got named according to the schema before it was saved. End-to-end demo, one prompt.
Flashcards — Applied
02 · Applied
Save-as-Skill
tap to reveal →
A conversational shortcut that promotes a one-off action you just performed in chat into a permanent, named skill the agent will run on future matching utterances.
← tap to flip back
02 · Applied
Naming Schema
tap to reveal →
A structured rule for file names — for example, Month-Year-Team-Purpose. Saved as a skill, it audits existing files and guides naming of new ones automatically.
← tap to flip back
02 · Applied
File Naming Audit Skill
tap to reveal →
A skill that checks files against a naming schema, flagging non-compliant names. Can fire on review or be expanded to trigger on file creation as a prevention step.
← tap to flip back
02 · Applied
Trigger Scope
tap to reveal →
The set of situations in which a skill fires — review, create, on-utterance, scheduled. Editing the scope is independent of editing the skill's content.
← tap to flip back
02 · Applied
Skills Reference (Markdown)
tap to reveal →
A Markdown document the agent generates listing every installed skill, its purpose, and when it fires. Perfect for handoff or documentation — produced from chat in one prompt.