Implement Bootstrap Case Runner

[jharris1679]

Overview

Implement bootstrap cases that apply common coding tasks to the user's actual repository. These test practical agent utility on real work.

Bootstrap Cases

ID	Task	Detection	Validation
boot-001	Fix linting errors	Run linter, find files with errors	Lint passes, tests pass
boot-002	Rename symbol	Find symbol to rename	Grep confirms, tests pass
boot-003	Extract duplicated code	Find duplicate patterns	Pattern gone, tests pass
boot-004	Add type annotations	Find untyped code	Type-checker passes
boot-005	Update deprecated API	Find deprecated usage	No deprecated patterns

Tasks

Detection System

• Implement detector interface
• Lint error detector (eslint, ruff, etc.)
• Duplicate code detector (basic pattern matching)
• Untyped code detector (for TS/Python)
• Deprecated API detector (configurable patterns)

Runtime Template System

• Template variable substitution (${LINT_ERRORS}, ${FILE_PATH})
• Dynamic case generation from detection results
• Skip case if nothing detected (no lint errors = skip boot-001)

Validation System

• "Minimal change" evaluator - penalize over-modification
• Non-regression check - tests must still pass
• Style adherence check - lint must pass after changes

CLI Integration

• sniff bootstrap command to run bootstrap cases
• Auto-detect which cases apply to current repo
• Show what was detected before running

Acceptance Criteria

• Bootstrap cases auto-detect issues in user's repo
• Template variables filled with detected items
• Agent attempts to fix detected issues
• Validation confirms fix without breaking tests
• "Minimal change" scoring penalizes over-engineering