The Agent Skills Directory

[COMMAND_EXECUTION]: The helper script scripts/convert.sh contains a command injection vulnerability. It uses the eval command to execute a string constructed from unvalidated input variables (INPUT_FILE and OUTPUT_FILE) within the execute_conversion function. Although the script wraps parameters in single quotes, it fails to escape single quotes within the input itself. An attacker could provide a file name containing a single quote followed by shell metacharacters (e.g., ';id;') to break out of the quoting context and execute arbitrary shell commands on the system.\n- [PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection as it processes untrusted document content without proper safeguards.\n
Ingestion points: The skill reads and converts various document formats including Markdown, HTML, DOCX, and CSV which may contain embedded instructions.\n
Boundary markers: No delimiters or protective instructions are provided to the agent to prevent it from obeying commands hidden within the ingested document content.\n
Capability inventory: The skill has the ability to execute shell commands (via the vulnerable script) and write files to the local file system.\n
Sanitization: There is no evidence of sanitization or validation of the input file content to detect or remove malicious instructions before processing.

doc-converter