guides
Findings
A finding is a proven, exploitable issue — not a maybe. Every one carries the agent's assessment, the exact evidence, and the fix. This page covers where findings live, how to read one, and how to triage it.
what you'll learn
- Where findings live and how to open the view
- How to read a finding: severity, the agent assessment, and evidence
- The four statuses and how to triage with one click
- How to filter, search, and confirm a fix
Where findings live
Findings are per organization. Open an org and click the alerts tab in its header — the full URL is /orgs/<your-org>/alerts. There is no cross-organization list; you always work inside one org at a time, so a finding can never leak between tenants.
Anatomy of a finding
Each row shows the essentials; click it to expand the full detail. A finding contains:
- Severity — one of six levels (below). Critical and high are what to fix first.
- Agent assessment — the ✦ agent assessment box: a verdict (confirmed, likely false positive, or uncertain), a confidence percentage, a plain-English explanation, and a
fix:line with the change that closes it. - Description — the full write-up of the issue.
- Evidence — a malicious input example (for code findings) or a code snippet, and, for issues that touch several places, an affected locations list of URLs or
file:linereferences. - Provenance — the footer line:
source severity(what the engine originally rated it, before Muster's adjustment) plusfirst seenandlast seen.
Severity levels
| parameter | type | description |
|---|---|---|
critical | severity | Immediate, high-impact exploit. Fix now. |
high | severity | Serious and exploitable. Fix in this cycle. |
medium | severity | Real risk with meaningful conditions or lower impact. |
low | severity | Minor or hard-to-exploit; fix opportunistically. |
info | severity | Informational — context or hardening, not a vulnerability. |
gas | severity | Gas — a smart-contract efficiency finding; surfaces only when scanning contracts. |
Status lifecycle
A finding is always in exactly one of four statuses. Muster opens it; you drive it to a resolution.
Open
newly proven, unaddressed
Acknowledged
triaged, in progress
Fixed
gone on re-scan
- Open — newly surfaced and unaddressed. The default for every new finding.
- Acknowledged — you've triaged it and work is underway.
- Fixed — set automatically when a re-scan no longer detects the finding (or mark it yourself). Muster verifies the fix on re-scan; it never changes your app or code. If it reappears on a later scan it reopens automatically (see Confirming a fix).
- False positive — not a real issue; set it aside without it counting against you.
Triaging a finding
- 1
Open the org's alerts view
Go to
/orgs/<your-org>/alertsvia the alerts tab. - 2
Narrow to what matters
Use the filter chips and search to focus — for example severity critical + status open. See Filtering & search.
- 3
Expand the finding
Click the row to open the assessment, description, evidence, and affected locations.
- 4
Set the status from the Triage column
Use the inline buttons — no dialog, one click:
Ack acknowledges · Mark fixed closes it (a re-scan does this automatically when the finding is gone) · FP marks it a false positive · Reopen returns a closed finding to open.
Triage many findings at once
Tick the checkbox on each row you want, or use the checkbox in the header to select the whole page. To go further, select all N matching selects every finding that matches the current search and filters, across pages. Then pick Acknowledge, Mark fixed, False positive, or Reopen from the action bar. Findings already at that status are left as they are, and the result tells you how many changed. Changing the search or filters clears the selection, so an action never touches rows you can't see.
Scripts and agents can do the same with the PATCH /orgs/:id/findings API endpoint or the bulk_update_findings MCP tool, up to 1,000 findings per call.
Send findings to GitHub issues
With GitHub connected, open a finding from the list and choose Create GitHub issue in the side panel, or select several findings and choose Create GitHub issues from the action bar. Each issue includes the severity, location and description, a link back to the finding, and the labels muster and severity:<level>. Findings that already have an issue show a GitHub #N link instead, and asking again returns the same issue. When the finding is fixed, whether you mark it or a re-scan confirms it's gone, Muster closes the issue for you.
Findings from code scans open issues in the repository they were found in. For web, DNS and cloud findings, choose a default repository under Settings → Integrations → GitHub. The Muster GitHub App needs the Issues: write permission; if GitHub asks an org admin to approve it, that's expected.
Issues are one of the GitHub actions you control, per repository, under Settings → Integrations → GitHub. They're off for new connections until you turn them on. While they're off for a repo, Muster won't open issues there and won't close ones it opened before. See Scanning → GitHub pull requests.
Filtering & search
The alerts view filters on three facets, each multi-select, plus a free-text search:
- Severity — critical, high, medium, low, info, gas.
- Status — open, acknowledged, fixed, false positive.
- Source — the engine that found it: code scanning, auditbase, web app, deps, cloud, DNS.
- Search — the Search alerts… box matches title, vulnerability type, file path, and description.
Add facets with + Add filter; each chip shows a live count, and clear all resets everything. A running n of total counter tells you how much the current filter is hiding.
Confirming a fix
Muster verifies fixes at the scan level, not with a per-finding button. After you ship the change, re-scan the asset — the agents re-run against the current state and a finding they can no longer reproduce stops recurring. If it ever comes back on a later run, it reappears with a regressed badge so a silent reintroduction never slips by.
Ways to re-scan
- From the dashboard — the asset's Scan button, or Scan all assets.
- From the API —
POST /api/v1/orgs/:id/scanwith{ "all": true }to re-test everything (handy in CI after a deploy).
Pull findings into your tools
Findings don't have to stay in the dashboard. The REST API and MCP server expose every field above — severity, the assessment, the fix, affected locations — so your AI coding agent can read a real finding while it writes the patch, then mark it fixed when it's done.