The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a CSP violation reporter endpoint (/api/csp-report) which acts as an ingestion point for untrusted external data.
Ingestion points: The violation object is parsed from req.body["csp-report"] in routes/csp-report.ts.
Boundary markers: None present for the logged output.
Capability inventory: The data is logged to the console and transmitted to AWS CloudWatch using the @aws-sdk/client-cloudwatch library.
Sanitization: The skill does not explicitly sanitize the fields from the incoming JSON report before logging or monitoring, which is a common surface for log injection, though appropriate for this skill's reporting purpose.
[EXTERNAL_DOWNLOADS]: The configuration whitelists well-known and trusted Content Delivery Networks and services.
The CSP policy includes https://cdn.jsdelivr.net, https://www.googletagmanager.com, https://fonts.googleapis.com, and https://fonts.gstatic.com to allow legitimate external resource loading.
[SAFE]: The skill correctly implements dynamic nonce generation for inline scripts and styles using crypto.randomBytes(16).toString("base64"), effectively mitigating XSS risks compared to using 'unsafe-inline'. It also provides a clear rollout strategy to prevent service disruption during security enforcement.

secure-headers-csp-builder