runtime-author

name	runtime-author
description	Author one runtime-neutral catalog asset that is adapter-translated per runtime (OpenCode/Claude) — single source, no duplicate per-runtime files.

runtime-author

Purpose

Every runtime asset (agent body, skill, workflow, command, template, hook) is authored once in the catalog as runtime-neutral markdown. The OpenCode adapter and the Claude Code adapter each translate it for their runtime at export time. This skill ensures the source is written so both adapters produce valid output — preserving single source (R-006) and the byte-parity moat.

When to use

When creating or modifying any catalog asset that will be rendered to a live runtime: agents, skills, templates, commands, workflows, hooks.

Procedure

1. Determine the asset kind (agent, skill, template, command, workflow, hook) and load its authoring standard from the catalog.
2. Author runtime-neutral. Write one source: frontmatter with every required field, body sections per the kind's standard. No runtime-specific syntax — the adapters translate the superset and drop what a runtime can't use.
3. Structural validity. Run aspis validate-runtime (agents) and resolve every broken reference before continuing.
4. Adapter render check. Run aspis export --dry-run; confirm no render errors for OpenCode or Claude Code. Fix the catalog — never patch the live file.
5. Byte-parity proof. Run aspis export --check; the catalog must regenerate the live runtime byte-for-byte. DRIFT means a hand-edit or a non-deterministic render — refuse the change until CLEAN.
6. Cross-runtime spot check. For the Claude Code adapter, verify the permission: block survives translation (known past asymmetry — a single stripped field is the worst kind of small change).

Outputs

• One runtime-neutral catalog asset, frontmatter + body, wired into its owner.
• A validate-runtime pass and an --dry-run render with no errors.
• A byte-parity verdict of CLEAN — both runtimes regenerate from one source.

Anti-patterns

• Authoring a separate file per runtime — defeats single source and R-006.
• Runtime-specific syntax in the catalog (e.g. tools: vs permission:) — breaks the other adapter's render.
• Skipping --dry-run or --check; trusting one adapter without testing both.
• Editing the live runtime file (.opencode/, .claude/) directly instead of the catalog source — guarantees the next parity check fails.
• Treating MISSING as DRIFT: a first export of a new asset is not drift and must not block the pipeline.