gspdev-prompt-audit

Auditing prompt files for semantic quality involves analyzing them to detect dead weight, contradictions, one-off patches, vagueness, and repetition, then generating a structured report to guide remediation.

Cross-file prompt analysis surfaces global patterns, inconsistencies, and duplication across skills and agents to enable consolidation. You need it when coherent AI behavior across multiple prompts is required.

Generate a prompt audit report by running an analysis scoped to agents or specific named files, which enforces baseline checks and per-file semantic criteria, writing the detailed output to dev/reports/prompt-audit.md.

Yes, you can limit the audit scope to skills, agents, or specific named prompt files to target the analysis and produce a focused semantic quality report for those exact files.

AI output degradation often stems from semantic quality issues in prompt files such as vague instructions, contradictions, or redundant dead weight that conflict with the model's default behavior.

The best way to detect contradictions across multiple prompt files is performing a cross-file analysis that flags inconsistencies across skills and agents to ensure coherent behavior and surface duplication.