StructKit Skill

What you get

• Prefer StructKit-native tools and MCP when available.
• Inspect structures and variables before generation.
• Preview diffs before writing files; avoid overwrites unless explicitly requested.
• Validate .structkit.yaml files and generated outputs.
• Keep custom structures reusable, composable, and CI-checked.

Source repository: httpdss/structkit-skills.

Install via skills.sh

The repository is laid out as a single root-level skill, so Skills-compatible installers can install it directly:

npx skills add httpdss/structkit-skills

If your installer accepts raw SKILL.md URLs, use:

npx skills add https://raw.githubusercontent.com/httpdss/structkit-skills/main/SKILL.md

Install via Hermes

hermes skills install https://raw.githubusercontent.com/httpdss/structkit-skills/main/SKILL.md --name structkit-workflows

Or register the repository as a tap (Hermes versions that support GitHub skill taps):

hermes skills tap add httpdss/structkit-skills

What's in the skill

.
├── SKILL.md
├── references/
│   ├── overview.md
│   ├── safe-generation.md
│   ├── authoring-structures.md
│   ├── ci-validation.md
│   └── publishing-structures.md
├── templates/
│   ├── basic-structure.structkit.yaml
│   └── repo-scaffold.structkit.yaml
├── scripts/
│   └── validate_structkit_repo.py
└── .github/workflows/validate.yml

• references/ — task-specific guidance the agent loads on demand (safe generation, authoring, CI, publishing).
• templates/ — starter .structkit.yaml files an agent can copy and adapt.
• scripts/validate_structkit_repo.py — validates a repo that hosts StructKit templates.
• .github/workflows/validate.yml — drop-in CI to keep custom structures honest.

Validate locally

Run the bundled validator against a directory of StructKit templates:

python scripts/validate_structkit_repo.py .

If structkit is installed on the same machine, the script also runs structkit validate against the templates it finds.

Pairing with MCP

The skill assumes the StructKit MCP server is reachable when possible. Start it alongside your AI assistant so the skill's workflows (list → inspect → preview → generate → validate) can use first-class tools instead of shelling out:

structkit mcp --server --transport stdio

See the MCP / AI Agent Workflow page for the matching client configuration and end-to-end tool calls.