The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process and analyze untrusted user-uploaded data, which is an inherent attack surface for indirect prompt injection.
Ingestion points: Processes multipart/form-data, S3/cloud storage uploads, and various file formats (images, archives, PDFs) provided by users.
Boundary markers: The skill explicitly instructs the use of sanitization tools (e.g., DOMPurify) and parser hardening (e.g., ImageMagick policies) to handle untrusted content.
Capability inventory: The skill utilizes Read, Write, Edit, and Bash tools to perform analysis and generate findings.
Sanitization: Implements a 12-step defense checklist including extension whitelisting, magic byte validation, AV/CDR scanning, and archive bomb protection.
[COMMAND_EXECUTION]: The skill uses the Bash tool to interact with an internal appsec-sdk for workflow orchestration.
Evidence: Section 7.1 contains shell commands such as appsec-sdk overlay.activate "<release-tag>" file_upload and mentions appsec-sdk finding.add. These commands are part of the intended integration with the vendor's security platform.

security-app-file-upload