What problem does it solve? Coding tasks often fail from unclear scope, missing verification, or lost progress across sessions. This Skill structures implementation and maintenance work around the target repository's contract so changes stay bounded, verified, and resumable. ## Core Features & Use Cases - Scoped execution: Infers the write set and minimum proof from the current diff, affected callers, and repository contract without requiring the user to supply file paths or test commands. - Proportionate verification: Matches evidence depth to risk, from focused checks for tiny changes to rollback proof for security, data, or deployment work, and distinguishes repo_verified, filesystem_projected, host_loaded, and live_accepted layers. - Resumable progress and executor selection: Recovers goal, completed work, and outstanding items on continue/resume, and selects GPT/Codex or GLM/ZCode executors by task fit rather than a fixed pipeline. - Use Case: Ask the agent to fix a bug in an API endpoint; it inspects the affected code, implements the change, runs the relevant existing tests, and reports exactly which verification layer was reached. ## Quick Start Use the ai-coding-workflow skill to implement and verify this coding task in the current repository.
