What problem does it solve? Determining whether a photo or video is authentic, original, and correctly captioned is difficult because most deceptive media is real footage with a false caption, and forensic filters like error level analysis are routinely misread. This Skill provides an ordered verification methodology that prioritizes provenance over pixel forensics and produces defensible, graded findings instead of unsupported verdicts. ## Core Features & Use Cases - Provenance-first verification: Reverse image search, keyframe extraction for video, archive corroboration, and C2PA Content Credentials checks to find the earliest copy and its original caption. - Physical and geometric consistency analysis: Shadow convergence, lighting direction, reflections, vanishing points, and optical signatures that hold up under scrutiny and survive recompression. - Signal-level forensics with honest limits: Clone detection, noise residuals, JPEG quantisation tables, double-compression detection, and clear guidance on when these tests are invalid (platform re-encodes, screenshots). - AI-generation and deepfake assessment: Durable structural tells versus tells that age badly, plus the documented failure modes of automated detector tools. - Use Case: A video circulates claiming to show an explosion in a named city. Extract keyframes, reverse search them, find the same footage published eighteen months earlier in a different country, detect an audio splice via spectrogram, and report it as confirmed recontextualised with graded confidence. ## Quick Start Verify whether this attached photo is authentic and correctly captioned, and give me a graded findings report with sources.
