WorkBuddy Configuration Health Audit

Problem

WorkBuddy configuration failures are often not caused by a single missing field, but by interactions among file paths, model context, safety rules, and credential handling. This skill evaluates the current configuration against an empty-configuration baseline: it does not ask users to match a template, but checks whether the setup is more likely than an empty config to cause lockups, leakage, overreach, or degraded execution.

How It Works

• Reads SOUL.md, IDENTITY.md, USER.md, AGENTS.md, models.json, mcp.json, and references/ files one by one; immediately redacts apiKey, token, password, and similar fields in models.json and mcp.json.
• Runs quick checks, baseline comparison, constraint detection, Pareto boundary analysis, and computes correctness, efficiency, and overall scores.
• Stops deeper analysis when three P0 violations are found; trend records store daily scores, ie, and violation counts, and flags significant degradation when scores drop by more than 10 points.

Boundaries and Notes

• Reports only issues with a clear causal chain; C9 content-attribution findings are marked as preliminary automatic screening and should be confirmed manually.
• Repair is disabled by default, limited to .md and .json whitelisted files, and every change requires per-item confirmation plus automatic backup.
• Deep checks depend on model instruction-following and context capacity; weaker models may produce larger deviation.