Agent Skill Blueprint
A reusable structure for authoring well-scoped context-engineering skills.
The template makes ownership boundaries, activation conditions, adjacent-skill exclusions, references, limitations, and claim provenance explicit, so major scope and attribution risks are visible. It does not specify concrete permissions, sensitive-data handling, user confirmation, external side effects, rollback, or data-flow controls, so substantial points are deducted.
The structure and authoring rules are broadly self-consistent and include actionable guidance, examples, and gotcha sections. However, the body is mostly placeholders, the linked references are stubs, and there are no reproduced key paths, abnormal-input behaviors, or diagnosable failure messages for a real skill; the static calibration therefore limits the score.
The template clearly targets creation of context-engineering Agent Skills and requires audience, activation triggers, and non-fit boundaries for neighboring skills. Its actual capability boundaries, input/output contract, platform fit, Chinese-language support, and mainland-China network reachability are unspecified, so points are deducted.
The information architecture is strong, covering progressive disclosure, practical guidance, examples, guidelines, gotchas, integration, references, and metadata; the repository context also provides an MIT license. The template leaves author, date, version, maintenance ownership, changelog, update path, and reference replacement unresolved, so it is not fully maintainable as delivered.
As an authoring scaffold, it can help produce a skill with clearer triggers, content, examples, and maintenance metadata. It is not itself a directly usable task-completion skill, requires substantial filling and validation, and provides no evidence of marginal benefit over manual authoring, so the score is low.
The file contains checkable structural requirements such as a body-length limit, metadata, examples, references, and claim IDs; repository context shows CI and tests, but those do not verify this template's key behavior as a skill. There is no third-party execution evidence or independent reproduction, so only limited credit is warranted.
- This is an unfilled authoring template and should not be installed or treated as an executable skill.
- The linked reference files are stubs and must be replaced or removed when creating a real skill, with auditable sources added.
- The template does not define concrete permission, sensitive-data, external-call, confirmation, or rollback controls; these must be designed for each resulting skill.
- Repository-level tests, benchmarks, and performance claims in the README do not demonstrate that this template itself passed those validations.
What does this skill do, and when should you use it?
This is a Markdown template for creating new Agent Skills in a context-engineering collection. It defines sections for discovery, activation boundaries, core concepts, detailed topics, practical guidance, examples, guidelines, gotchas, integration, references, and metadata. The template emphasizes explicit ownership boundaries, progressive disclosure, and moving detailed material into a references/ directory. It is useful for authors who want a consistent skill-authoring workflow, but it does not provide a concrete domain capability by itself.
Provides a SKILL.md outline with YAML frontmatter for name and description; guides authors through activation rules, adjacent-skill exclusions, concepts, detailed topics, executable workflows, examples, verifiable guidelines, failure modes, integrations, references, and version metadata; and recommends keeping the skill body under 500 lines while moving detailed material into references/.
- A maintainer adding a new context-engineering skill to the collection needs a consistent document structure.
- An author needs to define activation boundaries and prevent overlap with adjacent skills.
- A maintainer wants to separate detailed reference material from the primary skill body.
- A technical writer needs a checklist for examples, operational guidance, and experience-based gotchas.
What are this skill's strengths and limitations?
- Covers the major sections from discovery and activation through integration, references, and version metadata.
- Makes ownership boundaries, progressive disclosure, and executable guidance explicit.
- Requires examples, checkable guidelines, and gotchas, encouraging practical skill content.
- It contains no concrete context-engineering method or executable script.
- The body is a placeholder skeleton; authors must supply the topics, examples, references, and metadata.
- The source provides no independent test results or dedicated installation procedure for the template skill.
How do you install this skill?
The README documents directory-based installation rather than flattening SKILL.md into a single file. Copy the target skill directory into a host skill directory such as .codex/skills/ or .agents/skills/. The source does not document a template-specific installation command; collection-level Claude Code installation uses the marketplace workflow described in the README.
How do you use this skill?
Use template/ as the starting directory for a new skill, then fill in its name, description, activation conditions, ownership boundaries, guidance, examples, gotchas, references, and metadata. Example trigger: Create an Agent Skill for a context-engineering task and define when it activates, which adjacent skills it excludes, and the concrete workflow it teaches. The template requires third-person descriptions.