Commit and PR scans
Install the GitHub Action to scan on commit or pull request and return findings while the code is still in review.
ZeroQuarry combines multi-agent vulnerability research with the workflows security teams already use: commits, pull requests, scheduled jobs, disclosure inboxes, and patch review.
The red-team agent hunts for exploit chains and working proofs. The vendor-team agent tries to disprove each claim, challenge impact, and narrow remediation. That debate acts as guardrails against hallucinated or over-scored findings.
When the first pass is not enough, tell the agents to continue the debate. Push on a premise, ask for a better PoC, request an alternative patch, or keep triage running until the evidence is strong enough for your security bar.
Install the GitHub Action to scan on commit or pull request and return findings while the code is still in review.
Let ZeroQuarry draft narrow patches for confirmed issues, ready for engineering review instead of another ticket handoff.
Install the bot to suggest pull requests directly, with context from the finding and the remediation debate attached.
Run periodically and efficiently scan what changed since the last pull, so repeated coverage does not mean repeated waste.
Use lower-cost models to fan out over repositories, endpoints, binaries, and suspicious paths.
Promote only promising candidates into adversarial validation and evidence checking.
Spend stronger models on exploit construction, impact analysis, and patch quality.
When you receive a vulnerability report, forward the email to ZeroQuarry. The system analyzes the disclosure, identifies relevant repos or live targets, runs appropriate scans, and helps separate valid reports from vague or non-reproducible claims.
For outbound work, keep a history of disclosures you have made: affected product, timeline, status, evidence, PoCs, and communications. The same agent loop that validates findings can help prepare clearer disclosure packages.