agentleFS
volantlabsvolantlabs/vellis/.agents/skills/sysml-reference/SKILL.md

sysml-reference

Skill9 starsChanged 4 months agoApache-2.0View on GitHub

Ground textual SysML v2 and KerML authoring, review, language-semantics decisions, construct selection, and official-validator diagnosis in the repository's checksum-pinned specification corpus. Use for questions or changes involving parts, references, features, actions, bindings, successions, flows, ports, interfaces, states, constraints, requirements, views, imports, multiplicities, specialization, or other SysML/KerML syntax and semantics.

We split this skill into the rules it gives an agent

12 rules. About 0 are common, 6 are distinctive, and 6 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 6

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use the committed page corpus as a searchable projection of the pinned official PDFs.

    line 8
  2. Very specific:

    Do not replace this first step with a broad rg over pages/.

    line 21
  3. Very specific:

    If one specification is clearly in scope, rerun with --specification sysml-2.0 or --specification kerml-1.0 before broadening the search.

    line 25
  4. Distinctive:

    If results are broad or use different terminology, search outline headings before raw page text:

    line 27
  5. Very specific:

    Use raw rg over pages/ only for exact phrases, identifiers, or a cited section number.

    line 33
  6. Distinctive:

    Follow normative cross-references between SysML and KerML.

    line 34
  7. Distinctive:

    After changing a model, run the narrow scoped check while editing and the official validation gate required by the repository before …

    line 47
  8. Very specific:

    Before answering, verify that you can report the finder query (or the exact user-supplied citation), the smallest sufficient page set, each …

    line 50
  9. Very specific:

    Never edit files under reference/specifications/ manually.

    line 57
  10. Very specific:

    Use just model-reference-check to prove that Markdown, outlines, indexes, and manifests match the pinned PDFs.

    line 59
  11. Distinctive:

    Do not infer section semantics merely from an outline title.

    line 61
  12. Distinctive:

    Do not treat successful parsing as proof that a modeling interpretation is correct.

    line 63

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

SysML Reference

Use the committed page corpus as a searchable projection of the pinned official PDFs. Treat the PDFs as authoritative and the Markdown as generated retrieval material.

Workflow

  1. Unless the request already supplies an exact specification section or page, the first corpus command must be the ranked, outline-aware finder. Pass the modeling question or a concise description, not just one generic word:

    just model-reference-find "<question or construct description>"

    Do not replace this first step with a broad rg over pages/. The finder is designed to handle natural wording without loading every textual match.

  2. Review the ranked section titles and snippets. Open the best page, then read only the adjacent pages needed to finish the section. The finder routes to evidence; its rank is not an answer. If one specification is clearly in scope, rerun with --specification sysml-2.0 or --specification kerml-1.0 before broadening the search.
  3. If results are broad or use different terminology, search outline headings before raw page text:

    rg -n -i '<construct or section phrase>' reference/specifications/{sysml-2.0,kerml-1.0}/index.md

    Use raw rg over pages/ only for exact phrases, identifiers, or a cited section number.

  4. Follow normative cross-references between SysML and KerML. Search by both the construct name and any referenced section number, and prefer the normative Clause 8 definition when Clause 7 is explicitly informative.
  5. Inspect the pinned PDF under .cache/sysml/formal/ when a table, figure, mathematical layout, indentation, or questionable extraction affects the conclusion. Run just model-setup if it is absent.
  6. Separate the result into:
    • normative SysML or KerML semantics supported by cited sections and pages;
    • informative examples in the specifications;
    • Vellis repository profile or modeling conventions;
    • explicit inference where the sources do not directly decide the question.
  7. State the reference basis for every consequential conclusion using specification, section, and printed or physical page. Never present training-memory recall as specification evidence.
  8. After changing a model, run the narrow scoped check while editing and the official validation gate required by the repository before completion.

Before answering, verify that you can report the finder query (or the exact user-supplied citation), the smallest sufficient page set, each source's normative or informative status, any Vellis convention used, and every inference. If the pages read are much broader than the pages cited, narrow the retrieval before answering.

Corpus Rules

  • Never edit files under reference/specifications/ manually. Regenerate them with just model-reference-render after an intentional pinned-specification update.
  • Use just model-reference-check to prove that Markdown, outlines, indexes, and manifests match the pinned PDFs.
  • Do not infer section semantics merely from an outline title. Read the source text and its stated cross-references.
  • Do not treat successful parsing as proof that a modeling interpretation is correct; the official validator and the specification provide different evidence and both are required.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.