The cross-harness, self-improving memory layer for AI agents. Join our community: https://discord.com/invite/zdNChS2fBu
Go agent-ai-cliagent-harnessesagent-securitybrowser-extensionciclaude-codecloudcodexcursordetection-engineeringjevknowledge-basememorymobile-device-managementobservabilitysecuritysecurity-information-and-event-managementsecurity-toolstelemetrytraces
---
title: "Findings"
description: "Security risks, reusable workflows and habits Beacon Cloud notices in your sessions"
---
Every hour, Beacon Cloud reviews the new activity your connected devices forwarded and surfaces
what is worth your attention as findings. They are on the **Findings** page at
[beacon.sh/dashboard/findings](https://beacon.sh/dashboard/findings).
Findings are separate from the local [`beacon scan`](/cli/scan) results, which run open
[threat rules](/detections/index) on the machine and never leave it.
## Kinds of findings
| Kind | What it covers | Examples |
| --- | --- | --- |
| **Security** | Risky things your agents did | A secret pasted into a prompt or command, a destructive or privilege-raising command, an agent bypassing its approvals, prompt injection arriving through a tool result, a call to an unknown MCP server |
| **Skill** | Workflows worth reusing | Steps you walk an agent through more than once, drafted as a reusable Agent Skill |
| **Behavior** | Patterns in how you work with agents | Retry loops, abandoned sessions, approvals you always grant, expensive models used for simple tasks |
Each finding shows what happened, how severe it is, which sessions and harnesses it came from,
and a recommendation. A skill finding has a **Download skill** button that saves the drafted
`SKILL.md`, and findings carry a prompt you can copy into your agent to fix the problem.
## Review a finding
- Filter the list by kind (security, skills, behavior) and severity. Findings that still need
review come first.
- Open a finding to read its detail, the sessions it cites and the events it rests on. On
[Beacon Cloud Pro](/deployment/plans) the full evidence trail is shown.
- Use **Mark as** to set a finding open, resolved or dismissed.
- Mark it helpful or not helpful and add a note. The next analysis reads your feedback, so it
stops repeating what you rejected.
A finding that recurs updates the one you already have instead of creating a duplicate.
## Your first findings
Beacon Cloud analyzes a new account soon after its first sessions arrive, including the 30 days of
history a device uploads when it [connects](/cloud/devices). You get one email with your first
findings right after that analysis. Later findings arrive in a daily digest on the free plan, or
as real-time email alerts on Pro.
## Plan limits
| Plan | Findings |
| --- | --- |
| Beacon Cloud | 50 over the lifetime of the account |
| Beacon Cloud Pro | Unlimited |
| Enterprise | Unlimited, org-wide |
Every finding counts toward the free allowance, whatever its status. Once a free account has used
it, Beacon Cloud stops analyzing its new activity; its sessions keep uploading and stay
searchable. After an upgrade, findings resume from the next new telemetry. See
[Plans and pricing](/deployment/plans).
## Findings from your agent
On Beacon Cloud Pro, [Beacon Cloud MCP](/deployment/managed-mcp) lets the agent you are working in
list your findings (`beacon_findings_list`), read one with its evidence (`beacon_get_finding`) and
record your feedback or status (`beacon_finding_feedback`, which needs write access). Both
`beacon_get_finding` and `beacon_finding_feedback` accept a finding's dashboard URL.
## How analysis handles your data
- Only activity from devices you connected is analyzed, in an isolated sandbox per account.
- Events are projected to a compact, redacted record before analysis, so a secret is reported from
Beacon's redaction marker rather than its value.
- Every finding must cite events from your own sessions.
- On a device forwarding in Metadata-only mode, prompts, responses, tool arguments and output never
reach Beacon Cloud, so findings can only use what was sent. See
[What leaves your machine](/deployment/managed#what-leaves-your-machine).
## Related
<Columns cols={2}>
<Card title="Plans and pricing" icon="tag" href="/deployment/plans">
Free and Pro findings limits.
</Card>
<Card title="Sessions" icon="list" href="/cloud/sessions">
Open the sessions a finding cites.
</Card>
<Card title="Beacon Cloud MCP" icon="plug" href="/deployment/managed-mcp">
Read and review findings from your agent.
</Card>
<Card title="Local detections" icon="shield" href="/detections/index">
Offline threat rules with `beacon scan`.
</Card>
</Columns>