From research question to evidence-backed finding
Anvil brings application context, experiments, agent work, and reporting into one workspace. Built by Daniel Draghici (danieloizo) to keep complex security investigations connected and reviewable.
Private · invite-only. Access is limited to approved accounts.
About the researcher · Sign inExplore Anvil
Investigate
Organize related requests and inspect their responses in the same workbench.
HTTP Forge
Request tabs, named request contexts, and response status stay visible together.
Live requests · sample product data. The selected search shows a 200 response while two unselected missing-product tabs retain 404 feedback.
Explore
Move from captured traffic to the endpoints and context that frame the next question.
Traffic Atlas
Inspect the recorded requests behind a route instead of treating the inventory as a flat list.
Sample workspace data. Expanded DummyJSON routes expose their recorded observations and response status codes.
Coordinate
Track review tasks and inspect agent discussions alongside the research context.
Plans and tasks
A task list keeps completed work and next steps easy to inspect.
Completed documentation tasks. Three Done tasks record the capture, comparison, and documentation work behind the product review.
Agent discussions
Use bounded discussions to compare the supplied evidence, question a conclusion, and prepare a handoff.
Illustrative agent configuration · sample data. Claude Mythos 5.1 and GPT-6 Sol illustrate model selection, with Daybreak enabled for GPT-6 Sol. The sample review distinguishes observed fields from conclusions that still need validation.
Review
Bring review records, their source context, and report drafts together.
Finding review
Keep the observation, expected behavior, and proposed follow-up beside the review status.
Illustrative finding · awaiting review. An illustrative UI-validation proposal keeps its context and review notes in one view.
Reports
Carry the observed result and its source trail into a report.
Observed sample-data review. The product review keeps its observed result, source trail, and decision together.
Built by Daniel Draghici (danieloizo)
Anvil is the private AI security research harness built and used by Daniel Draghici, a professional security researcher known as danieloizo.
View Bugcrowd profile · Discuss a research engagement
For application security research inquiries, contact Daniel: contact [at] anvilsec.eu.
Operated by Dialog Soft SRL. Responsible use and disclosure contact.