Feature: auto-detect codegen/schema issues from real usage and (opt-in) auto-file GitHub issues

[cultron]

Motivation

Real-world usage surfaces CLI/codegen defects that pass validate_schema but fail at runtime — e.g. #59 (uuid PK generated with no generation strategy → every insert fails). Catching these today requires a human to notice the runtime failure, diagnose it, build a minimal repro, and file manually. That loop is slow, and most users won't do it — so defects go unreported and unfixed.

Request

An opt-in capability (a CLI "skill", also exposed as an MCP tool) that detects when the CLI produces inaccurate or poor output during normal commands and can automatically open a well-formed GitHub issue against apsoai/cli, grounded in the real usage that triggered it — with user consent.

Detectable-signal examples

• A generated entity that cannot satisfy its own constraints on insert (NOT NULL PK with no generation strategy / no default — exactly primaryKeyType: "uuid" generates a non-generated @PrimaryColumn() → every insert fails with null id #59).
• validate_schema passes but apso dev / runtime errors (validation ↔ runtime divergence).
• Codegen emitting unused imports, dead decorators, or asymmetric branches (e.g. serial generates a PK strategy but uuid doesn't).
• Migration diffs that fail against the PGlite sandbox.
• Schema features silently ignored (e.g. nullable dropped on enum fields).
• Repeated runtime errors correlated with a specific generated artifact.

Behavior / guardrails

• Opt-in (--report-issues flag or config) — schemas/data may be sensitive.
• Sanitize/redact schema + data; include only a minimal repro, versions, and the offending command/artifact.
• Dedupe against existing issues (search first; comment rather than duplicate).
• Explicit confirmation before opening — especially via the MCP tool.
• Auto-attach environment (apso --version, OS, node), the relevant .apsorc excerpt, and the generated snippet.

Integration

• A CLI command (e.g. apso doctor --report) and an MCP tool (report_issue / diagnose) so AI agents driving apso can auto-file with the user's OK.

Value

Closes the loop between real usage and CLI quality. Defects like #59 would be detected and filed automatically from actual usage instead of relying on a user to notice, diagnose, and hand-write the report.