Secret scanning that tells you which leaked keys are still live.
ARGUS secret scanning finds hardcoded credentials three ways, filters out the usual noise, and labels each secret live, rotated, test or unknown. You rotate the keys that can still be used, first, instead of working through a list of placeholders.
Three passes, one list
Each scan runs three detection passes, then removes the matches that are almost never real secrets.
- 01
Vendor patterns
Known key formats for cloud providers, payment platforms and SaaS APIs, using the open gitleaks rules.
- 02
High-entropy strings
Random-looking values that match no known pattern but behave like keys or tokens.
- 03
Named assignments
A literal value assigned to a variable named like a credential, such as api_key or client_secret.
- 04
Noise filtered out
Test fixtures, documentation placeholders, publishable keys, environment variable references, commit hashes and example config files.
From scan to fix
- 01
Scan the repository
On every full scan, and on the changed lines of each pull request.
- 02
Three passes run
Patterns, entropy and assignments, each catching what the others miss.
- 03
Validity is labelled
With AI scanning on, each secret is marked live, rotated, test or unknown.
- 04
Rotate, then gate
Start with live keys. A gate policy can fail a build on new secret findings.
Built for the rotation, not the report
Live keys first
Validity labels separate a working production key from a revoked one or a test value.
Caught in the pull request
A key pasted into a change is flagged on that line, before it reaches your main branch.
Ranked with everything else
Secrets share one priority list with code, dependency and infrastructure findings.
What it does not do yet
We would rather you hear it here than find it in week two.
- Validity labels need AI scanning turned on. Without it, secrets are still detected but not labelled.
- Very large files are skipped, and each pass reports a capped number of findings per scan.
- Rotating a key is still your step. Removing a secret from code does not remove it from git history.
Secret scanning questions, answered
Straight answers about Secret scanning and how ARGUS does it. Updated 17 November 2026.
Still have questions? Contact us
What is secret scanning?
Secret scanning searches code and configuration for credentials that should not be there, such as API keys, access tokens, passwords and private keys. ARGUS scans with three passes (vendor patterns, entropy and named assignments), filters common noise and labels each secret live, rotated, test or unknown.
Why are hardcoded secrets dangerous?
Anyone who can read the repository, or a leaked copy of it, can use the credential. Secrets also stay in git history after the line is deleted. MITRE tracks the weakness as CWE-798, use of hard-coded credentials. Finding them early limits how far a leaked key can travel.
How does ARGUS find hardcoded secrets?
Three passes run on every scan: known vendor key patterns using open gitleaks rules, high-entropy strings that look like keys or tokens, and literal values assigned to credential-named variables such as api_key or client_secret. The results are combined into one list.
What do live, rotated and test mean in ARGUS?
With AI scanning turned on, each secret gets a validity label. Live means it still looks usable, rotated means it has been replaced, test means a test or sample value, and unknown means it could not be decided. Rotate live keys first.
What should I do when a secret is found?
Revoke or rotate the credential with its provider first, because removing it from code does not make it safe. Then move it to a secrets manager or environment variable, and check the provider's logs for use you do not recognise. ARGUS ranks live secrets first so you start in the right place.
Does deleting a secret from the code fix the leak?
No. The value remains in git history and in every clone and fork made before you deleted it. Treat any committed secret as exposed: rotate it, then clean it up. Deleting the line only stops it spreading further.
How does ARGUS reduce secret scanning false positives?
It filters out matches that are almost never real secrets: test fixtures, documentation placeholders, publishable keys, environment variable references, commit hashes, integrity hashes, lockfiles and example or template config files. Admins can also turn individual rules off.
What is the difference between secret scanning and secrets management?
Secret scanning finds credentials that leaked into code. Secrets management stores credentials securely and injects them at runtime, so they never need to be in code. You need both: management prevents leaks, scanning catches the ones that happen anyway.
Does ARGUS catch secrets in pull requests?
Yes. Secret detection runs on the changed lines of every ARGUS pull request review, so a key pasted into a change is flagged on that line before it reaches your main branch.
Is secret scanning part of every ARGUS scan?
Yes. Secret detection runs on every full repository scan and on the changed lines of every pull request review. Validity labels (live, rotated, test) need AI scanning turned on; without it, secrets are still found but not labelled.
The rest of ARGUS, on the same ranked list
ARGUS reviews every pull request for security: inline comments with a fix, a merge gate you control, and fix pull requests a person approves.
ExploreARGUS featureSASTARGUS SAST pairs rule-based static analysis with AI passes for taint flow, business logic, LLM apps and race conditions.
ExploreARGUS featureSCAARGUS dependency scanning matches your lockfiles to OSV advisories, marks each CVE reachable or not with evidence, ranks by EPSS and CISA KEV, and blocks malicious packages..
ExploreARGUS featureSBOMARGUS generates an SBOM on every scan in CycloneDX 1.6 and SPDX 2.3, adds VEX from reachability, and builds infrastructure, crypto and post-quantum BOMs..
ExploreSee all of ARGUS. Further reading: How to find hardcoded secrets in code · Securing CI/CD pipelines. Standards: OWASP Secrets Management Cheat Sheet · CWE-798: Use of hard-coded credentials.
Try it on one repository.
Connect a GitHub repo and see the findings, ranked, in your first scan.