| name | aalc-automation-practices |
| description | Apply AALC-specific conventions when implementing, debugging, or reviewing desktop automation, image or template matching, task orchestration, cooperative stopping, PySide6 UI, configuration, debug switches, or code structure in AhabAssistantLimbusCompany. |
AALC Automation Practices
Before editing, read the target signature, callers, and similar repository implementations. Do not impose generic software-design preferences on desktop automation behavior that has been tuned against a real UI.
Load by task
- For
auto.*, ImageUtils.*, OCR, UI location, retries, or automation flow, read references/automation-design.md.
- For task threads, start/stop, emulator waits, PySide6 UI, language/theme, or config, read references/lifecycle-ui.md.
- For replaying a user screenshot, load
replay-matching and its production-pipeline reference.
- Before adding or changing
debug_*, also read .opencode/tools/debug_model_constitution.md.
Load only the matching reference.
Core workflow
- Establish the real failure state, UI hierarchy, and call chain.
- Inspect API defaults, search regions, template cropping, and similar code.
- Choose the smallest change that explains the root cause; do not stack parameters as trial and error.
- Keep the call graph flat unless a realistic reuse or semantic boundary justifies an abstraction.
- Run the most relevant syntax, lint, test, or replay verification.
- Review whether the diff is more complex than the failure and remove redundancy introduced by the change.
Constraints
- Do not re-instantiate project singletons.
- Do not use user
config.yaml as a template or synchronization target.
- Do not replace cooperative stopping with forced thread termination.
- Do not present general image-matching heuristics as facts without replay evidence.
- Do not change unrelated legacy code.
