PyPI · prowler
Prowler: Stored XSS in HTML reports through unescaped cloud resource tags
Prowler's HTML output formatter inserts finding.resource_tags into the generated report without HTML escaping. A cloud principal who can create or edit a resource tag in an account that is later scanned can store HTML or JavaScript in that tag. When another user opens the generated Prowler HTML report, the payload executes in the report page.
This is stored cross-site scripting in a security report artifact. The attacker needs permission to influence tags on a scanned cloud resource, and the victim must open the generated HTML report. In that context, JavaScript can read and modify the report DOM, alter displayed findings, and interact with any same-origin local or hosted report content available to the browser. This can undermine trust in generated security findings and can expose data contained in the report page.
The repository has 13,875 GitHub stars, which is above the 10,000 star floor. SECURITY.md is present and does not reject GitHub private reporting. The dedicated GitHub private vulnerability reporting endpoint for prowler-cloud/prowler returned true.
The vulnerable sink is present in current master at commit 329dfdf8e6cb8bc0424fb54b6595408e20969782 and in the latest GitHub release tag 5.28.0. The package metadata identifies the Python package as prowler.
The HTML formatter builds each finding row with an f-string in prowler/lib/outputs/html/html.py. Neighboring fields are explicitly escaped:
resource_uid is escaped at line 85 with .replace("<", "<").replace(">", ">").status_extended is escaped at line 87 with the same pattern.The tag column does not apply equivalent escaping:
<td>{parse_html_string(unroll_dict(finding.resource_tags))}</td>
unroll_dict concatenates tag keys and values into a string, and parse_html_string only prefixes each item with an HTML bullet. Neither function escapes <, >, quotes, or event-handler attributes before insertion into the table cell. As a result, a tag value such as this remains active markup in the generated report:
<img src=x onerror="window.PROWLER_TAG_XSS=1">
<img src=x onerror="window.PROWLER_TAG_XSS=1">
A proof HTML fragment using the same formatter structure set window.PROWLER_TAG_XSS=1 from resource_tags, while equivalent payloads in resource_uid and status_extended stayed text because those fields are escaped.
Local evidence:
/home/unkn0wn/security_audit/vulnerability-reports/prowler-html-tag-xss-evidence-2026-05-26.json/home/unkn0wn/security_audit/vulnerability-reports/prowler-html-tag-xss-proof-full.htmlThe minimized generated row below mirrors the vulnerable formatter output. The first and third candidate payloads are escaped controls. The middle payload is the unescaped resource_tags value and executes when the report is rendered.
<tr class="table-danger">
<td><img src=x onerror="window.PROWLER_UID_XSS=1"></td>
<td>
•Owner=<img src=x onerror="window.PROWLER_TAG_XSS=1;document.body.setAttribute('data-prowler-tag-xss','1')">
</td>
<td><img src=x onerror="window.PROWLER_STATUS_XSS=1"></td>
</tr>
Browser proof result from the local artifact:
{"tag": 1, "uid": 0, "status": 0, "attr": "1"}
Escape tag keys and values before inserting them into HTML output. Prefer a single HTML escaping helper for every finding field that can originate from provider data. If bullets or separators are needed, build the list structure after escaping each key and value, rather than escaping the final HTML string after markup has been added.
Is your project exposed to this? Stateward checks every dependency on every pull request and flags it only if your code actually reaches it.
Check my repoSources: CISA KEV (public domain), OSV.dev & GitHub Advisory Database (CC-BY-4.0), FIRST EPSS, NVD/CWE (public domain). Served live from the Stateward advisory database.