| name | subsystem-hardening-orchestration |
| description | Use for coordinating a LawnBerry Pi hardening pass across high-risk subsystems such as navigation, mission recovery, manual control/camera, and AI quality. Covers routing to the right hardening skill, sequencing focused changes, and keeping validation and docs aligned. |
| argument-hint | Which subsystem should be hardened: navigation, mission recovery, control/camera, or AI? |
| user-invocable | true |
Subsystem Hardening Orchestration
What this skill does
This skill routes hardening work to the right specialized pass so risk-heavy subsystems get focused treatment instead of generic maintenance.
When to use
- the task is a hardening pass rather than a one-off bug fix
- you need to improve runtime credibility or safety for a known subsystem
- the work spans code changes, regression coverage, and maintainer documentation
Routing guide
Choose the leading pass based on the subsystem:
- navigation behavior, movement confirmation, stop/fault handling ->
navigation-hardening-pass
- mission persistence, restart, pause/resume recovery ->
mission-recovery-pass
- manual control, RoboHAT USB handoff, joystick responsiveness, camera streaming ->
control-camera-regression-review
- AI quality improvement behind the current API ->
ai-model-quality-pass
Procedure
- Start with
maintainer-reentry.
- Choose one primary hardening pass; do not mix several unless evidence demands it.
- Perform a focused defect or risk audit before editing.
- Make the smallest credible hardening changes for that subsystem.
- Extend targeted regression coverage close to the seam.
- Apply
maintainer-doc-sync if maturity, limitations, or interfaces changed.
- End with a concise statement of what was hardened versus what remains field-limited.
Guardrails
- avoid turning a hardening pass into broad feature work
- keep evidence and validation close to the claimed improvement
- prefer deterministic failure handling over optimistic state advancement
Completion checks
- one subsystem led the pass
- focused regression coverage exists or was explicitly planned
- maintainer docs reflect any changed maturity or runtime behavior
- remaining limitations are stated plainly