AccessForge is an open-source workspace for accessibility journey authoring and evidence-oriented review. Define user goals, independent success criteria, permitted origins and bounded actions; manage access and inspect desktop-runner readiness. Built with Codex. This preview demonstrates the deployed workspace and authoring flow; real VoiceOver/NVDA baseline, repair and independent rerun evidence are not yet demonstrated.
What real task does your product handle with GPT-6 Astra?
Maker
AccessForge uses GPT-6 Astra through a local Codex CLI adapter to propose accessibility diagnoses, scoped source repairs and bounded navigation actions from explicitly authorized inputs. Structured output is validated before a proposal can advance; human approval and independent success criteria remain separate from the model’s own claims. The repository pins the model profile to gpt-6-astra and keeps ChatGPT authentication with Codex rather than copying credentials into the service. The submitted preview demonstrates the deployed workspace, journey authoring, permission boundaries and runner readiness. It does not demonstrate a completed real VoiceOver/NVDA baseline, applied repair or independent rerun; full execution-host acceptance remains pending. Source: https://github.com/gnanam1990/ac...
Report
Maker
📌
Hi Product Hunt! I built AccessForge to make accessibility work traceable: what was authorized, what a person wanted to achieve, and what independent evidence would prove success. The current preview includes project and environment setup, immutable journey authoring, role-based access, usage boundaries and runner-readiness views. It is not a claim of a completed screen-reader test: the demo shows authoring and administration, with no qualified desktop runner enrolled. The next milestone is one real form-error recovery journey, a scoped repair and an independent rerun. Source: https://github.com/gnanam1990/ac.... I would love feedback from accessibility engineers on the journey definition and evidence workflow.