Audit AI-produced work with structured sections on confidence and shortcuts.
npx skills add https://github.com/kategage/progressive-ai-skills --skill confess
Please help me install this Agent Skill. Skill: confess Source: https://github.com/kategage/progressive-ai-skills/tree/main/skills/confess Command: npx skills add https://github.com/kategage/progressive-ai-skills --skill confess
Honest, unvarnished evaluations of AI-produced work help users identify shortcuts, hidden assumptions, and weak spots, improving reliability and decision-making.
After generating a substantial output, trigger a confession by saying confess.
High-intent search queries and answers about installing and using this skill.
A self-review audit reveals confidence levels, shortcuts taken, unverified assumptions, and known weaknesses in AI-generated work. It uses a structured confession format to expose hidden flaws in significant outputs like strategy analyses or drafted documents.
The best way to check AI confidence on a strategy analysis is to trigger a structured confession. This generates a Confidence Inventory and highlights the weakest parts of the document for transparent error-detection.
Yes, you can apply this transparency check to technical or medical documents. The confession template is domain-agnostic and can be adapted to different contexts by adjusting the framing and categories for legal, medical, or technical use cases.
Perform an honest self-audit by triggering the confession command after generating the research brief. This produces a structured report detailing shortcuts, unverified assumptions, and things the AI noticed but did not handle.
When checking AI shortcuts and hacks, the limitation is that the audit relies on the AI's own internal awareness. It flags unverified assumptions and weak spots but does not independently verify external facts or validate data beyond its generated context.
lexijamesesqRefine strategy documents through iterative section-level edits and voice alignment.
AGM1968Identify and tag assumptions in plans with verification statuses.
sanky369Generate SEO-optimized articles with keyword research and on-page optimization.
human-avatarClassify source independence and assess evidence convergence for claims.
