The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The scripts/ink-verify.py script employs the standard xml.etree.ElementTree module for SVG parsing. This module is susceptible to XML entity expansion (billion laughs) attacks, which could lead to resource exhaustion if a malicious SVG is processed.
[INDIRECT_PROMPT_INJECTION]: The pdf_info function in scripts/ink-verify.py performs unconstrained decompression of PDF streams using zlib.decompress(). This creates a vulnerability to decompression bombs (Zip bombs) that can exhaust memory and cause a denial of service when the agent processes a crafted PDF.
[INDIRECT_PROMPT_INJECTION]: The skill provides an attack surface for indirect prompt injection because it ingests untrusted file data and possesses file-write and CLI execution capabilities.
Ingestion points: SVG, PDF, and PNG file verification and processing in scripts/ink-verify.py and the Bash scripts.
Boundary markers: No specific boundary markers are used to isolate file content from the agent's instructions.
Capability inventory: Writing files to the filesystem and executing Inkscape command-line actions.
Sanitization: There is basic shell delimiter filtering in the Bash scripts, but the Python logic lacks security-hardened parsing for complex file formats.
[COMMAND_EXECUTION]: The skill facilitates the execution of shell scripts and the Inkscape binary on the host system. While this is the intended functionality, it constitutes a broad attack surface that depends on the integrity of the local environment and the Inkscape binary path.

inkscape