| name | harness-config-patterns |
| description | Use when working in the harness-config repository on implementation structure, projection architecture, resource and dir composition, target override precedence, profile and ignore integration, or code placement. |
Harness config Patterns
Purpose
Use this skill for implementation design in this repository. The spec defines
the required behavior; this skill helps place the code and avoid projection
mistakes.
References
references/code-map.md: source files and ownership by behavior.
references/projection-architecture.md: activation pipeline, path bases,
target overrides, profiles, ignores, and composition.
references/antipatterns.md: common implementation mistakes.
Workflow
- Decide whether the task changes the standard contract. If it does, use
$harness-config-specs first.
- Read the narrowest reference that locates the behavior.
- Keep behavior in public core APIs before formatting it in the CLI.
- Use
$harness-config-testing for the regression shape.
Guardrails
- Keep physical source paths, logical source paths, output-relative paths, and
target output paths distinct.
- Preserve the one-way projection model: runtime folders are outputs.
- Do not infer undeclared targets from existing folders.
- Do not edit generated
.agents or .claude files as source.