The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Installs the @cyberdrk/gram package from the NPM registry during the installation phase.
[DATA_EXFILTRATION]: Accesses sensitive authentication data, including Instagram session cookies (sessionid, csrftoken, ds_user_id) and local browser profile directories (e.g., Chrome, Firefox, Safari) to retrieve authentication tokens.
[COMMAND_EXECUTION]: Executes shell commands using the gram binary to perform authenticated actions such as viewing posts, following users, and posting comments.
[PROMPT_INJECTION]: Ingests untrusted external data from Instagram (posts, comments, search results) which represents a surface for indirect prompt injection.
Ingestion points: Commands such as gram comments, gram feed, and gram search retrieve content from Instagram's APIs.
Boundary markers: The instructions do not define specific boundary markers or warnings to the agent regarding embedded instructions in retrieved data.
Capability inventory: The skill possesses write capabilities including gram comment, gram follow, gram like, and gram save which could be abused if the agent follows instructions found in external content.
Sanitization: No evidence of sanitization or filtering of the retrieved Instagram content is provided in the skill documentation.

gram