Summary
Gitea: Fork-PR Actions task can read a third private repository via the collaborative-owner branch (missing fork-PR guard)
GetActionsUserRepoPermission (models/perm/access/repo_permission.go) decides whether an Actions
task token may access a target repo. Its cross-repo branches each enforce a fork-PR discriminator ,
except the collaborative-owner branch, which is missing the !task.IsForkPullRequest guard that
its sibling has. As a result, when a private repo B lists owner A as a collaborative owner, an
attacker-controlled fork pull-request workflow whose base repo is owned by A is granted code-read
on B, i.e. the fork's YAML can clone a third private repository it has no rights to.
Details
// models/perm/access/repo_permission.go (v1.26.2), in GetActionsUserRepoPermission
if checkSameOwnerCrossRepoAccess(ctx, taskRepo, repo, task.IsForkPullRequest) { // passes isForkPR -> denies forks
return maxPerm, nil
}
...
if taskRepo.IsPrivate { // <-- NO IsForkPullRequest check here
actionsUnit := repo.MustGetUnit(ctx, unit.TypeActions)
if actionsUnit.ActionsConfig().IsCollaborativeOwner(taskRepo.OwnerID) {
return maxPerm, nil // grants code-read to target repo B
}
}
The sibling same-owner path correctly denies fork PRs:
func checkSameOwnerCrossRepoAccess(ctx, taskRepo, targetRepo, isForkPR bool) bool {
if isForkPR {
return false // Fork PRs are never allowed cross-repo access to other private repositories.
}
...
}
taskRepo = the repo whose workflow is running (the PR's base repo A); repo = the target being
cloned (B). IsCollaborativeOwner(taskRepo.OwnerID) asks "does target B's Actions config trust A's
owner for cross-repo read?" When B trusts ownerA, the branch returns maxPerm (code-read) even whentask.IsForkPullRequest is true, i.e. when the executing YAML is the fork's, not A's.
Every sibling enforces the fork-PR discriminator; except for this branch:checkSameOwnerCrossRepoAccess denies forks; ComputeTaskTokenPermissions
(models/actions/token_permissions.go) only clamps the token ceiling to read-only for fork/cross-repo
(its own comment notes the access decision is in GetActionsUserRepoPermission, so it does not
neutralize the gap, it just makes the leak read-only); secrets (models/secret/secret.go) and the
approval gate (services/actions/notifier_helper.go) both correctly key on IsForkPullRequest.
Reachability, the runner clones target repo B over git-HTTP with the task token:routers/web/repo/githttp.go → GetDoerRepoPermission(ctx, repoB, ActionsUser) →GetActionsUserRepoPermission(ctx, repoB, actionsUser, taskID) with IsForkPullRequest == true →
collaborative-owner branch returns code-read → p.CanAccess(Read, code) passes → private clone of B
succeeds. (CheckRepoScopedToken in githttp is a no-op for the Actions token.)
PoC
Setup: private base repo A (usera/repoA), private third repo
B (userb/repoB) with a planted SECRET.txt, B's Actions config trusting usera as a collaborative
owner, and a genuine running fork-PR task token (token_hash computed with Gitea's own HashToken)
presented as HTTP Basic. Requesting GET /userb/repoB.git/info/refs?service=git-upload-pack:
| Condition (same fork-PR token) | HTTP | Meaning |
|---|---|---|
| anonymous (no token) | 401 | auth required |
| token, A public, B trusts A | 404 | branch gated on taskRepo.IsPrivate ⇒ A public skips it |
| token, A private, B has no collab-owner config | 404 | no trust ⇒ denied |
| token, A private, B trusts A (collab-owner) | 200 | git clone of private B succeeds |
| config removed / restored | 404 / 200 | deterministic |
In the 200 case, git clone of private repo B succeeded and yielded its SECRET.txt, the full source
of a third private repo the fork-PR author has no rights to.
Suggested remediation
Add the same fork-PR guard the sibling path has (one line):
if taskRepo.IsPrivate && !task.IsForkPullRequest {
actionsUnit := repo.MustGetUnit(ctx, unit.TypeActions)
if actionsUnit.ActionsConfig().IsCollaborativeOwner(taskRepo.OwnerID) {
return maxPerm, nil
}
}
This flips Vuln_ForkPR_LeaksThirdPrivateRepo to PASS, keeps Control_NonFork_Allowed PASS
(legitimate collaborative-owner sharing still works), and leaves the existingTestGetActionsUserRepoPermission suite all green.
Impact
Read-only confidentiality breach: discloses the full source of a third private repository (B) to an
untrusted external fork-PR author. Read-only, not write/RCE.
Preconditions (honest):
- B is deliberately configured with a collaborative owner, but that is exactly the feature's intended
use, so realistic for any deployment using it. - The fork PR's base repo A is itself private (the branch is gated on
taskRepo.IsPrivate). Forking a
private A already requires read on A, so this is a normal internal-contributor situation, not a
weakening, the escalation is "read A (granted) → read a different private repo B (never granted)." - The fork-PR workflow must actually run, most realistically via an attacker who had one earlier PR
approved (the "approved before" path inifNeedApproval), after which fork PRs auto-run.
The application does not correctly enforce access controls, allowing a principal to access resources or operations beyond their granted permissions. Typical impact: unauthorized data access or execution of privileged operations.
CVE-2026-58416 has a CVSS score of 6.3 (Medium). The vector is network-reachable, low privileges required, and no user interaction. A CVSS score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether this affects your application depends on whether the vulnerable code is present and reachable in your environment. A fixed version is available (1.27.0); upgrading removes the vulnerable code path.
Affected versions
Security releases
Kodem intelligence
Severity tells you how bad this could be in the worst case. It does not tell you whether you are exposed. Exploitability and impact are functions of runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A vulnerable package can sit in your dependency tree and never run.
Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter. Kodem's runtime-powered SCA identifies whether this CVE is reachable in your applications.
Already deployed Kodem?
See it in your environmentNew to Kodem? Get a demo →Remediation advice
Kodem Kai can prioritize this vulnerability in your dependency tree and generate a fix recommendation.
Frequently Asked Questions
- What is CVE-2026-58416? CVE-2026-58416 is a medium-severity incorrect authorization vulnerability in gitea.dev (go), affecting versions < 1.27.0. It is fixed in 1.27.0. The application does not correctly enforce access controls, allowing a principal to access resources or operations beyond their granted permissions.
- How severe is CVE-2026-58416? CVE-2026-58416 has a CVSS score of 6.3 (Medium). This score reflects the worst-case severity of the vulnerability, not your specific exposure. Whether it represents real risk in your environment depends on whether the vulnerable code is present and reachable.
- Which versions of gitea.dev are affected by CVE-2026-58416? gitea.dev (go) versions < 1.27.0 is affected.
- Is there a fix for CVE-2026-58416? Yes. CVE-2026-58416 is fixed in 1.27.0. Upgrade to this version or later.
- Is CVE-2026-58416 exploitable, and should I be worried? Whether CVE-2026-58416 is exploitable in your environment depends on whether the vulnerable code is present and reachable. A CVSS score is a worst-case rating; it does not account for your specific deployment, configuration, or usage patterns. Kodem, an Intelligent Application Security platform, uses runtime intelligence to show which vulnerabilities actually execute in production, so you can focus on the ones that represent real risk. Get a demo
- What actually determines whether CVE-2026-58416 is exploitable, and how bad it is? Exploitability and impact are not fixed properties of a CVE. They depend on runtime truth: whether the vulnerable code is present, reachable, and actually executes in your application. A high CVSS score on a dependency that never runs is not the same as real risk. Kodem, an Intelligent Application Security platform, uses runtime intelligence to reveal which vulnerabilities actually execute in production, so teams prioritize the ones that genuinely matter.
- How do I fix CVE-2026-58416? Upgrade
gitea.devto 1.27.0 or later.