The Agent Skills Directory

[Prompt Injection] (LOW): The /tweets style and /tweets feature commands allow for arbitrary string input. These could be leveraged for prompt injection to manipulate agent behavior, although the scope is limited to text generation styling.
[Data Exposure & Exfiltration] (LOW): Several commands, such as /tweets project <path> and /tweets changelog <path>, accept arbitrary file paths. This introduces a potential path traversal vulnerability where the agent might be prompted to access sensitive local files if input validation is not enforced by the underlying implementation.
[Indirect Prompt Injection] (LOW): The skill's primary function involves processing untrusted data from external project files.
Ingestion points: Reads content from CHANGELOG.md, README.md, and package.json at user-defined paths.
Boundary markers: No explicit boundary markers or 'ignore embedded instructions' warnings are documented in the skill definition.
Capability inventory: File read operations are performed on the specified paths.
Sanitization: No sanitization or validation of the file content is mentioned prior to interpolation into the generation prompt.

tweet-ideas