@sylang/skills
v0.1.0
Published
Hermes agent skills for Sylang — 9 description-based-loading markdown skills covering the full DSL (23 file types, ~44 relationship keywords). Drop into ~/.hermes/skills/ or the equivalent skill directory of any Claude-Code-style agent.
Readme
@sylang/skills
Hermes agent skills for Sylang. 9 description-based-loading markdown files covering every Sylang file type (23 of them) and the cross-file traceability rules that bind them together.
Install
pnpm add @sylang/skillsThen copy the markdown files into your agent's skills directory:
# On the agent host:
cp -R node_modules/@sylang/skills/skills/sylang ~/.hermes/skills/Or symlink, or commit-vendor into your agent repo — whatever fits your deployment model. The skills are plain markdown with frontmatter; no build step.
What's included
| File | Covers |
|------|--------|
| 00-fundamentals.md | DSL grammar (hdef/def/ref), the 23 file extensions, ~44 relationship keywords, project layout, naming conventions. Read this first. |
| requirements.md | .req .haz .sgl .sam .itm — requirements, hazards, safety goals, safety analysis, item definition (ISO 26262) |
| features-variants.md | .fml .vml .vcf .fun — feature models, variant selection, variant configs, function networks |
| architecture.md | .blk .ifc .smd .ucd .seq — blocks, interfaces, state machines, use cases, sequence diagrams |
| safety-analysis.md | .flr .fta — FMEA AIAG-VDA and Fault Tree Analysis |
| tests-product-line.md | .tst .ple .sylangextend .spr — test cases, product-line root, DSL extensions, sprint planning |
| spec-dash.md | .spec .dash — reactive documentation and live dashboards |
| agents.md | .agt — Hermes agent definitions written in Sylang |
| traceability.md | Cross-file relationship rules + coverage analysis workflows |
Why 9 files instead of 23 or 1?
- One file per extension would duplicate the shared DSL grammar 23 times.
- One mega-file would bloat every agent turn's context.
- Domain-grouped (this layout) matches how the file types are actually
used together: a hazard analysis (
.haz) is never written without the matching safety goal (.sgl) and requirement (.req) files, so they share a skill.
Loading model
These skills are designed for description-based loading (Claude Code
style). Each file's frontmatter description field lists every file
extension it covers verbatim, so when the agent encounters a .fml file
it grep-matches the skill list and loads features-variants.md —
plus the always-needed 00-fundamentals.md.
If your agent loads all skills unconditionally, total cost is ~25 KB markdown / ~6500 tokens. Acceptable for a Sylang-focused agent; avoidable on a general-purpose one by leaning on the description-match selection.
License
Apache-2.0. See LICENSE.
Author
Balaji Boominathan (@balaji-embedcentrum)
