Build a Concise Custom Skill

What you will complete

A validated skill folder with clear activation, minimal resources, and no unsafe default.

Make the operating decision

Choose a short action-oriented name and write a description that says what the skill does and when it applies. Put activation language in metadata. Keep SKILL.md focused on purpose, workflow, important constraints, and direct links. Use references for schemas or mode-specific guidance and scripts only when deterministic repeated logic improves reliability.

Define two explicit invocation requests, two realistic implicit requests, and one or two near misses. A near miss matters because an overly broad skill can hijack unrelated work. Keep automatic discovery unless an explicit-only policy was actually requested. Do not infer explicit-only use merely because the workflow has approval gates.

Run scripts on representative inputs, run the current skill validator, inspect interface metadata, and scan for absolute personal paths, credentials, unfinished scaffold text, and undocumented writes. Structural validation does not replace behavioural testing, but an invalid structure should be repaired before model evaluation.

Worked fictional example: Harbourlight Home Services

The fictional estimate-exception skill activates for requests to build the weekly unbooked-estimate brief. It does not activate for writing a customer follow-up message. A small script validates CSV headers and creates stable row references; judgment about which exceptions matter remains in written business rules.

Harbourlight Home Services and every estimate record in this course are fictional. They demonstrate the method and do not represent a Flowgrammer client, a deployed system, or measured savings.

Complete workbook section 4

Use the evidence available for your own bounded job. Write unknown when the evidence is missing, and record the person or action that can resolve it.

1. Write the skill name, description, purpose, and stop boundary.
2. Create only required scripts, references, and assets.
3. Write explicit, implicit, and near-miss trigger cases.
4. Validate the structure and every deterministic script.

Open the course workbook PDF or read the workbook details.

Critical gate before continuing

• The description is discriminating.
• Conditional references are directly discoverable.
• Scripts have declared inputs and failures.
• The skill has no secret or external write default.

If a gate fails, repair the current section, narrow the scope, leave the route manual, or record a blocked or stop decision. Continuing is not the only successful learner action.

Common failure modes

• Expanding beyond a validated skill folder with clear activation, minimal resources, and no unsafe default. before the current artifact can be graded.
• Turning a missing value, unavailable source, or blocked integration into a confident conclusion.
• Treating a prompt instruction as proof that the effective tool or permission boundary works.
• Marking a manual, simulated, or untested route as live.

Check your application