Name the core abstraction and its failure modes.
Wish Course
Repository inspection and local conventions
How a coding agent learns the shape of a codebase before changing it.
Learning Objectives
What this lesson should make precise
Translate the concept into a Maple/Hermon proposal contract.
Define at least one evaluation case that can fail the model safely.
Tutorial Flow
How this lesson becomes a demo and training target
Each tutorial is written as a user education path and a model-improvement artifact. The diagram shows how the idea moves into a lab, a typed contract, an eval gate, and then a Hermon/MapleAI demo route.
Concept
Fast orientation
Applied Lab
Applied lab: Repository inspection and local conventions
Output Contract
files_to_inspect, commands_to_run, owned_changes, unrelated_changes_policy
Eval Gate
Does the answer separate model proposal from deterministic execution?
Demo Route
Wish training and public model checks
01
Fast orientation
A coding agent should inspect file lists, package manifests, test setup, framework conventions, and nearby implementation patterns before editing.
- Use file search.
- Read nearest code.
- Respect local style.
02
Dirty worktree protocol
If unrelated files are dirty, preserve them. If touched files contain user changes, read carefully and work with them rather than reverting.
- Do not reset.
- Do not clean.
- Do not overwrite user work.
03
Scope discipline
A high-quality patch is narrow, testable, and reversible. Unrelated refactors make review harder and hide risk.
- Patch only the cause.
- Avoid metadata churn.
- State residual risk.
Lab
Applied lab: Repository inspection and local conventions
Produce an inspection checklist for a repo where the user reports a broken demo chat frame.
Expected result
- A typed JSON-style proposal rather than free-form advice.
- Clear authority boundaries and denied operations.
- A test or rubric that decides whether the proposal is deployable.
Evaluation
How Maple would grade this work
Rubric
- Does the answer expose assumptions instead of hiding them?
- Does the answer separate model proposal from deterministic execution?
- Does the answer produce artifacts that can be tested, reviewed, and rolled back?
Output contract
files_to_inspect, commands_to_run, owned_changes, unrelated_changes_policyUse this lesson as training direction
A strong lesson gives users a mental model and gives Hermon a sharper target for examples, probes, and demo prompts.