<< All versions
Skill v1.0.0
currentAutomated scan100/100alysisai/alysiscode/skill-creator
──Details
PublishedOctober 21, 2026 at 09:44 PM
Content Hashsha256:3832eda3e7462930...
Git SHA8b19afb60340
──Files
Files (1 file, 1.8 KB)
SKILL.md1.8 KBactive
SKILL.md · 21 lines · 1.8 KB
version: "1.0.0" name: skill-creator description: Create or update a reusable Alysis Code skill through its supported scaffold and validation workflow.
Use when: The user wants to create or revise a reusable Alysis Code skill.
Do not use when: The user needs a one-off answer, repository change, or prompt that should not become a skill.
Treat commands below as suggestions; use the normal tool and approval flow.
- Interview only for missing decisions that materially affect the skill: target requests, trigger boundary, required workflow, authorization limits, and reusable resources.
- Inspect an existing skill before updating it. Preserve supported metadata and resources that still serve the requested workflow.
- For a new skill, use
alysis skill init <name>as the suggested scaffold so naming, paths, and frontmatter follow the same conventions as the CLI. Do not recreate scaffold logic in prose or ad hoc files. - Make the description concise and discriminating. Put essential model instructions in
SKILL.md; add scripts, references, or assets only when they provide a concrete reusable benefit. - Keep instructions imperative, scoped, and explicit about external actions. Do not assume tools or services unavailable in the target environment.
- Exercise any new script and test at least one realistic invocation when practical.
- Remove temporary test inputs, generated scratch files, and other throwaways only when created during the current task; preserve pre-existing and user-provided files. Generated workflows must use this ownership boundary, clean those throwaways before final verification, and rerun verification after any later cleanup or edit.
- Finish with the suggested command
alysis skill validate <skill-path>. Resolve validation errors and report warnings or remaining behavioral uncertainty.