The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The scripts/batch_generate.py script is vulnerable to path traversal. The id field from the input CSV is used directly to construct output file paths using os.path.join (lines 45-46) without sanitization, allowing an attacker to specify paths that write files outside the intended output directory.
[COMMAND_EXECUTION]: The scripts/batch_generate.py script uses subprocess.check_call to execute the companion script scripts/generate_qr.py. While the call uses an argument list to prevent shell injection, it executes commands with parameters derived from external CSV data.
[INDIRECT_PROMPT_INJECTION]: The skill's batch workflow in scripts/batch_generate.py processes external CSV data (URLs and labels) without boundary markers or specific isolation instructions, creating a surface where malicious instructions could be passed to the agent.
Ingestion points: Data is read from a CSV file path provided to scripts/batch_generate.py via the --csv argument (line 33).
Boundary markers: None identified for isolating or delimiting CSV input.
Capability inventory: Subprocess execution and local file system writes.
Sanitization: The skill implements URL validation and SVG text escaping, but lacks validation for the batch ID field used in file paths.

qr-code-generator