Skill Rails

Build one standalone skill from a small canonical source graph while leaving genuine meaning judgments with the AI or user.

Before making an authoring or maintenance judgment, read §§0–0.1 of references/skillEvolutionMethod.md. Use its routing table and the decision's uncertainty, reach, and reversal cost to choose any additional sections. The generated references/skillEvolutionMethod.index.json is only a byte-range navigation aid; it does not define meaning or a closed task taxonomy.

Work from canonical source

• Change the source package, target, module, observer, renderer, or contract that owns the behavior. Never hand-edit a generated target carrying .skill-rails-build.json.
• Put domain meaning in one canonical domain source, never in generated output or runtime state.
• Declare only the source modules, imports, inputs, output, contracts, and mechanism files the target actually consumes. The build fails closed on unknown fields, undeclared paths, root escapes, and unsupported combinations.
• Use exact IDs and paths. Do not infer requirement, check, effect, or causal relationships that the source graph does not declare.
• Treat missing behavioral or host evidence as unproven; build and integrity checks prove delivery only.
• Do not import archived Skill Rails or Devflow as an implementation baseline, compatibility layer, or runtime fallback.

Author or change one target