| name | hermes-skill-author |
| description | Authors Hermes-compatible SKILL.md workflows. |
| version | 1.0.0 |
| platforms | ["linux","macos","windows"] |
| metadata | {"hermes":{"tags":["hermes","skills","skill.md"],"category":"skills","requires_toolsets":["terminal"]}} |
Hermes Skill Author
When to Use
Use this skill when you need to create, review, or refactor a reusable Hermes Agent SKILL.md workflow for ZAI Coder or another local-first agent environment.
Procedure
- Identify the concrete workflow the skill should teach.
- Keep the frontmatter description short enough for catalog discovery.
- Use this section order:
When to Use, Procedure, Pitfalls, and Verification.
- Put supporting material under
references/, reusable outputs under templates/, helper code under scripts/, and supplementary files under assets/.
- Declare platform limits only when the workflow truly depends on an operating system.
- Declare required environment variable names, but never store secret values in the skill.
- Prefer read-only or planning behavior by default; require explicit approval for state-changing operations.
- Run the marketplace validator and security report before enabling the skill.
Pitfalls
- Do not invent commands that the target project does not expose.
- Do not load broad reference material into the main
SKILL.md when a focused reference file is enough.
- Do not request apply-level permissions such as
providers:apply or execution:apply unless the workflow genuinely mutates state.
- Do not erase source provenance when importing from a hub, GitHub path, well-known endpoint, or direct URL.
Verification
Confirm the skill is discoverable in the catalog, has a safe entrypoint, passes manifest validation, records source provenance, and only enables after the dry-run and approval gates pass.