The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill uses the command gh auth token to retrieve a sensitive authentication token and includes it in an HTTP header for external requests.
[DATA_EXFILTRATION]: The instructions command the agent to send the GitHub token to URLs parsed from untrusted Linear comments and descriptions. This allows an attacker to capture the token by providing a malicious link.
[COMMAND_EXECUTION]: The skill executes shell commands like curl and gh using variables derived from external, untrusted content without verification.
[PROMPT_INJECTION]: The skill processes untrusted data from Linear issue descriptions, comments, and external transcripts. It lacks boundary markers or sanitization, creating an attack surface for indirect prompt injection where malicious instructions in the data could hijack the agent's logic. (Ingestion points: Linear description and comments in mcp__linear; Boundary markers: None; Capability inventory: curl, gh, mcp__linear, Read tool; Sanitization: None)

linear-issue