The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructions in SKILL.md suggest installing the linkedin-api package from the public PyPI registry to handle API interactions.
[DATA_EXFILTRATION]: The skill utilizes sensitive LinkedIn session cookies (LINKEDIN_LI_AT and LINKEDIN_JSESSIONID) obtained from environment variables. These credentials are used exclusively to authenticate with LinkedIn's official domain to perform the requested actions.
[PROMPT_INJECTION]: The skill processes untrusted content from LinkedIn profiles, messages, and feed posts, which presents a surface for indirect prompt injection.
Ingestion points: The script scripts/lk.py fetches data using functions like get_conversations, get_feed_posts, and get_profile.
Boundary markers: Data is printed directly to the console without explicit delimiters or warnings to the agent to disregard instructions within the data.
Capability inventory: The script's functionality is limited to data retrieval and display; it does not perform file writes, network requests to third-party domains, or arbitrary command execution.
Sanitization: Fetched content is displayed as-is without filtering or sanitization of potential injection strings.

linkedin