WasItAIGenerated

Build me a focused AI detection across text, image, audio and video workflow for the personal core of WasItAIGenerated. Requirements:

- Use exactly this stack: Next.js 15 + TypeScript + SQLite. Model source documents, editable pages or notes, internal links, revisions, and exports; retain source IDs and timestamps.
- Paid product context: Detection API and tools for AI-generated text, images, audio and video, plus a thesis review. Build only this DIY scope: Analyze user-supplied text with a labelled heuristic or user-selected detector, show evidence and uncertainty, and export a reviewable report.
- For AI detection across text, image, audio and video, import a user-supplied document and preserve its original text and hash.
- Run a clearly labelled heuristic or user-selected detector and record its model/version, score, and evidence spans.
- Let the user review uncertain passages and export a report with limitations and source excerpts.
- Use a local web page with input, progress, review, and export views. Required input or access: OpenAI or Anthropic API key in .env; A README stating plainly that the output is a guess, not a detector, and must never be used to accuse anyone. Keep credentials in .env.
- Acceptance: with one labelled sample, show the input, saved intermediate state, and exported result; verify this invariant: A classifier score cannot be presented as proof of authorship; a failed detector run remains unknown and never overwrites source text.
- Out of scope: a trained classifier and the labelled corpus behind it; a measured false-positive rate you can quote to an institution. Keep this a personal, inspectable workflow.
- Include a README with setup, a sample input, required keys or permissions, data location, and the supported scope.

# AGENTS.md — Build guide for WasItAIGenerated

## Project scope
Build a private workspace that pastes in text, asks one general model for an AI-likelihood judgement with reasoning, stores results, and exports them.

Catalogue verdict: no. The interface is a text box; the product is a fine-tuned classifier. Detection quality does not come from a clever prompt. It comes from training a model on a labelled corpus and tuning it until the error rate is low enough to act on, which is months of work against a target that moves every time a new generator ships. Asking a general model "is this AI-written?" produces confident, unreliable answers, and in an academic-integrity setting that means falsely accusing real students. A one-shot build gets you the UI and none of the responsibility.
Use the implementation prompt below to define the deliverable. Complete each phase's acceptance checks before extending the scope.

## Working agreement
- Inspect the repository and its existing instructions before choosing paths, dependencies or commands. Keep one coherent stack and explain changes to the proposed architecture.
- Plan a vertical slice that accepts a real input and produces the useful output described below. Persist only the state the prompt calls for; respect memory-only and upstream-managed workflows. Use fixtures only when they are clearly labelled.
- After scaffolding, document the actual install, development, check and build commands in README and keep them synchronized with the package or project manifest. Do not report commands as successful unless they ran.
- Work in small steps. At handoff, list implemented flows, checks actually performed, remaining blockers, and any credentials or provider setup the owner must supply.
- Do not publish, spend money, contact customers, delete source data or run irreversible migrations without the project owner's authorization.

## Prerequisites
- OpenAI or Anthropic API key in .env
- Node.js 22
- SQLite database
- A README stating plainly that the output is a guess, not a detector, and must never be used to accuse anyone

## Stack and architecture
- Use exactly this stack: Next.js 15 + TypeScript + SQLite.
- Model source documents, editable pages or notes, internal links, revisions, and exports; retain source IDs and timestamps.
- Interface for this AI detection across text, image, audio and video workflow: a local web page with input, progress, review, and export views.

## Security and data integrity
- A classifier score cannot be presented as proof of authorship; a failed detector run remains unknown and never overwrites source text.
- Keep API keys and OAuth secrets in .env; redact them from logs, generated files, and error responses.
- State the practical limit: a trained classifier and the labelled corpus behind it.

## Agent implementation rules
- Project rule, data: Model source documents, editable pages or notes, internal links, revisions, and exports; retain source IDs and timestamps.
- Project rule, behavior: For AI detection across text, image, audio and video, import a user-supplied document and preserve its original text and hash.
- Project rule, recovery: A classifier score cannot be presented as proof of authorship; a failed detector run remains unknown and never overwrites source text.

## Optional agent skills and references
No installable skill is listed for this guide; use the project rules and the stack documentation.

Read the linked SKILL.md and its dependencies before adding a skill. Select only the skills matching this project's runtime and task; their documentation does not supply API access, credentials or approval to perform external actions. Pin the reviewed revision where the tool supports it. Follow the chosen agent's documented project-level installation mechanism.

## Distribution ideas
These are optional planning notes. Obtain the owner's approval before publishing or contacting anyone.
- Demonstrate this AI detection across text, image, audio and video operation with sample or user-owned data: For AI detection across text, image, audio and video, import a user-supplied document and preserve its original text and hash.
- Show a concrete result or export: Let the user review uncertain passages and export a report with limitations and source excerpts.
- State the practical limit: a trained classifier and the labelled corpus behind it.

## Engineering roadmap
1. Phase 1, architecture and data: Use exactly this stack: Next.js 15 + TypeScript + SQLite. Model source documents, editable pages or notes, internal links, revisions, and exports; retain source IDs and timestamps.
2. Phase 2, implement: For AI detection across text, image, audio and video, import a user-supplied document and preserve its original text and hash.
3. Phase 3, implement: Run a clearly labelled heuristic or user-selected detector and record its model/version, score, and evidence spans.
4. Phase 4, review and output: Let the user review uncertain passages and export a report with limitations and source excerpts.
5. Phase 5, recovery and acceptance: Verify this invariant with a saved fixture: A classifier score cannot be presented as proof of authorship; a failed detector run remains unknown and never overwrites source text. State the practical limit: a trained classifier and the labelled corpus behind it.

## Paid-product capabilities outside this build
- a trained classifier and the labelled corpus behind it
- a measured false-positive rate you can quote to an institution
- detection for images, audio and video, not just text
- per-sentence highlighting instead of one document-level guess
- retraining as generators change

## Implementation prompt
Build me a focused AI detection across text, image, audio and video workflow for the personal core of WasItAIGenerated. Requirements:

- Use exactly this stack: Next.js 15 + TypeScript + SQLite. Model source documents, editable pages or notes, internal links, revisions, and exports; retain source IDs and timestamps.
- Paid product context: Detection API and tools for AI-generated text, images, audio and video, plus a thesis review. Build only this DIY scope: Analyze user-supplied text with a labelled heuristic or user-selected detector, show evidence and uncertainty, and export a reviewable report.
- For AI detection across text, image, audio and video, import a user-supplied document and preserve its original text and hash.
- Run a clearly labelled heuristic or user-selected detector and record its model/version, score, and evidence spans.
- Let the user review uncertain passages and export a report with limitations and source excerpts.
- Use a local web page with input, progress, review, and export views. Required input or access: OpenAI or Anthropic API key in .env; A README stating plainly that the output is a guess, not a detector, and must never be used to accuse anyone. Keep credentials in .env.
- Acceptance: with one labelled sample, show the input, saved intermediate state, and exported result; verify this invariant: A classifier score cannot be presented as proof of authorship; a failed detector run remains unknown and never overwrites source text.
- Out of scope: a trained classifier and the labelled corpus behind it; a measured false-positive rate you can quote to an institution. Keep this a personal, inspectable workflow.
- Include a README with setup, a sample input, required keys or permissions, data location, and the supported scope.

## Completion evidence
Demonstrate the prompt's acceptance scenarios against the scoped workflow. Include setup from a clean checkout and failure recovery. Check persistence across restart and export/restore only for the state the prompt says to store; for memory-only tools, confirm that temporary content is discarded as specified. Record actual results and remaining limitations. A detailed plan alone does not establish a working replacement.