Add your source
Choose the application files you want Deadbolt to inspect.
Deadbolt's primary judge path is a keyless, evidence-backed Codex security audit. Its separate local API engine supports the hunt, patch, and re-analysis loop when configured.
PRIMARY PATH · NO API KEY
Run keyless in CodexAsk$deadbolt to audit source you own.KEYLESS IN CODEXREAD-ONLY AUDITEVIDENCE BACKED
Run keyless in CodexPRIMARY JUDGE PATH · CODEX SKILL
Run $deadbolt keylessly in Codex.
Install the Deadbolt Skill once, then audit the included InvoicePilot source without a separate API key.Read-only source reasoning audit · does not run the web engine or claim patch verification.01INSTALL02ASK
RUN KEYLESS IN CODEX$deadbolt03AUDIT SAMPLETHE MISSING SECURITY TEAM
Find the bug.Fix the risk.Re-analyze it.
Deadbolt gives solo builders the security loop enterprise teams already have: understand the application, find what matters, write the fix, and show whether the same hunt still finds it.
For repositories and deployments you own or are authorized to test.
The web product is the API-backed hunt → patch → re-analysis engine. The keyless $deadbolt Codex Skill is a separate read-only reasoning audit; it does not run this engine or claim patch verification.
BUILT FOR THE DEVELOPER WITHOUT
A SECURITY TEAMCODEQLCI/CDTIME TO BECOME AN EXPERT
ILLUSTRATIVE REMEDIATION FLOW · LOCAL API ENGINE
How the local engine is designed to work.Illustrative until a real run is recorded.
Confirm authorization
Only audit code you own or are explicitly allowed to test.
Run the full loop
Use the API-backed engine to review evidence, focused patches, and affected-hunt re-analysis.
ILLUSTRATIVE REMEDIATION FLOWNOT A RECORDED RUN
ILLUSTRATIVE ACCESS CONTROL EXAMPLE
Example: one signed-in user can read another user's private record.
This example shows the evidence and before→after framing the local API-backed engine is designed to produce. It is not a claim that a live patch or verification run has been recorded.
BEFORE · EXAMPLE CROSS-TENANT READ→AFTER · EXAMPLE 404 DENIED
Run the keyless auditTHE CORE LOOP
A finding is not finished until the patched clone is re-analyzed.
Every result connects evidence, human impact, a focused patch, and a fresh run of the affected hunt lens.
01MAP
Threat-model the app
Trace entry points, identity, sensitive data, and trust boundaries before judging risk.
02HUNT
Reason across the code
Four focused passes hunt secrets, auth flaws, injection, and dangerous configuration.
03PATCH
Write the smallest fix
Turn every confirmed finding into a focused, reviewable diff with the human stakes attached.
04RECHECK
Re-analyze the patch
Run the affected hunt lens again against the patched clone and show whether the original root cause remains.
THE SAFETY BOUNDARY
Defend what you own. Never weaponize what you find.
Deadbolt reports evidence, plain-English impact, and patches. It does not probe third-party targets or produce turnkey attack tools.