The Agent Skills Directory

[COMMAND_EXECUTION]: The bash script find_lib_source.sh is vulnerable to shell command injection. It accepts library names and versions as command-line arguments and interpolates them directly into shell commands like grep and sqlite3 without any sanitization or escaping. This allows a user or a malicious prompt to execute arbitrary shell commands by including metacharacters like semicolons, pipes, or backticks in the input.
[DATA_EXFILTRATION]: The skill provides tools to traverse the local filesystem, specifically targeting the Conan cache (~/.conan2) and local project directories. It queries the Conan SQLite database and reads the contents of conanfile.py files, which may expose sensitive information regarding project dependencies, build configurations, and internal code structures.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection by ingesting and processing data from local conanfile.py files and directory listings. If these files are part of an untrusted repository, they could contain malicious instructions designed to subvert the agent's behavior. * Ingestion points: Contents of local conanfile.py and directory structures. * Boundary markers: None present. * Capability inventory: Ability to execute shell commands and read local files. * Sanitization: The find_lib_source.sh script lacks any input validation or escaping for the processed arguments.

conan-source-lookup