Security guide · Updated
Dependency vulnerability triage: what to fix first
Dependency scanners report every known vulnerability in every package you install, and most of those alerts do not put your app at real risk. Triage means ranking them by evidence: is the flaw being exploited in the wild, how likely is exploitation, can your code reach the vulnerable function, and how bad would it be. Fix the few that pass those tests first.
By George Shvaya, founder of Webvello · Published
Key takeaways
- CISA’s Known Exploited Vulnerabilities catalog listed 1,781 vulnerabilities exploited in the wild as of November 16, 2026; anything on it that you run comes first.
- CISA revoked Binding Operational Directive 22-01 on June 10, 2026 and replaced it with BOD 26-04, which weighs KEV status alongside exposure, automation and impact.
- FIRST’s EPSS estimates the probability that a published CVE will be exploited in the wild in the next 30 days, updated daily.
- NIST says the National Vulnerability Database built a significant enrichment backlog from early 2024 that it could not clear, so many CVEs lack NVD severity data.
- The 2025 npm attacks, the debug account takeover and the Shai-Hulud worm that compromised more than 500 packages, came from malicious releases, not old CVEs.
Which signals should decide what to fix first?
| Signal | Question it answers | Source |
|---|---|---|
| CISA KEV | Is this vulnerability known to be exploited in the wild? | CISA |
| EPSS | How likely is exploitation in the next 30 days? | FIRST |
| CVSS | How severe is it if exploited? (v4.0 published November 2023) | FIRST |
| Reachability | Does your code actually call the vulnerable function? | GitHub Dependabot docs |
CISA calls the KEV catalog “the authoritative source of vulnerabilities that have been exploited in the wild,” and it showed 1,781 entries when we checked. EPSS, from FIRST, estimates “the probability that a published CVE will be exploited in the wild in the next 30 days.”
Reachability is the signal most teams lack. GitHub’s Dependabot can show whether vulnerable function calls are detected for JavaScript and TypeScript repositories (in public preview), but its docs add that “a call path does not prove that a vulnerability can be exploited.” Treat reachability as evidence, not proof, in either direction.
A practical triage order
- 1
Malicious or compromised releases
A package version published by an attacker is not a bug to schedule. Remove or pin away from it now and rotate any secrets the build could reach. - 2
In KEV and reachable or exposed
Exploited in the wild and in a path your app uses, or on a system facing the internet. Patch immediately. - 3
High EPSS or in KEV, reachability unknown
Patch soon, and check reachability while the fix is in review. - 4
High severity, low likelihood, not reachable
Schedule in normal dependency updates, and record why it waited. - 5
Dev-only or unused code paths
Dismiss with a recorded reason. Dependabot’s dismissal reasons include “not used” and “tolerable risk”; use them so the next person knows why.
Drowning in dependency alerts?
Send us your repository’s alert list. We will tell you which ones matter and which can wait.
What changed in 2026 for vulnerability prioritization?
| Change | What happened | Source |
|---|---|---|
| BOD 22-01 revoked | The 2021 directive that set KEV remediation deadlines for federal agencies was revoked on June 10, 2026 and superseded by BOD 26-04. | CISA |
| BOD 26-04 | “Prioritizing Security Updates Based on Risk”: KEV status is one of four urgency factors, with asset exposure, exploit automation and technical impact. | CISA |
| NVD backlog | NIST said it developed a significant backlog of unenriched CVEs starting in early 2024, could not clear it, and moved backlogged CVEs published before March 1, 2026 to “Not Scheduled”. | NIST |
The NVD change matters for small teams: a scanner that relies only on NVD severity scores will miss data for many recent CVEs. Use sources that aggregate advisories from each ecosystem, such as OSV.dev, which describes itself as “an aggregator of vulnerability databases” for open source, and the GitHub Advisory Database. BOD 26-04 binds federal agencies; for everyone else it is a useful model. It is also listed in our rules tracker.
What did the 2025 npm attacks change?
| Incident | What happened | Source |
|---|---|---|
| debug account takeover | “On 8 September 2025, the npm publishing account for debug was taken over after a phishing attack.” Version 4.4.2 was malicious; 4.4.3 is the fix. | GitHub advisory GHSA-4x49-vf9v-38px |
| Shai-Hulud worm | CISA said the worm “has compromised over 500 packages” and advised pinning to releases produced before September 16, 2025. | CISA alert, Sept 23, 2025 |
| npm response | GitHub was notified on September 14, 2025 and removed more than 500 compromised packages. | GitHub blog |
- Commit lockfiles and install with the lockfile, so a new malicious version cannot arrive on its own.
- Delay adopting brand-new versions of critical packages by a few days unless they fix a security issue.
- Keep publish credentials behind phishing-resistant two-factor authentication.
- Watch for advisories about malicious packages, not only CVEs, in the scanners you use.
Related services and resources
Pages that go deeper on the pieces this one touches.
Common questions
Straight answers to what owners ask us first.
Why do we get so many dependency alerts we cannot act on?
Scanners match package versions to advisories. They usually cannot tell whether your code calls the vulnerable function, whether the package only runs in development, or whether the flaw is being exploited. Most alerts fail at least one of those tests, which is why triage matters.
Is CVSS enough to prioritize vulnerabilities?
No. CVSS rates how severe a flaw would be if exploited. It does not say how likely exploitation is or whether your app is exposed. Combine it with KEV status, EPSS and reachability.
Does BOD 26-04 apply to private companies?
No. Binding operational directives apply to U.S. federal civilian agencies. Private organizations often follow them as a reference for good practice.
Can you triage our dependency alerts?
Yes. Our AI code and dependency audit sorts your open alerts by exploitation evidence and reachability, fixes or upgrades what matters, and records why the rest can wait.
Fix the alerts that matter
We triage dependency vulnerabilities by evidence and patch what puts your app at risk.