The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the bash tool to execute composio commands that incorporate user-provided $ARGUMENTS. Specifically, the pattern composio execute GORGIAS_GET_TICKET -d '{"ticket_id":"<ID>"}' where <ID> is replaced by user input is vulnerable to command injection if the input contains shell metacharacters like semicolons, backticks, or quotes to break out of the JSON string.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection.
Ingestion points: Untrusted data enters the agent context via the message threads retrieved by GORGIAS_GET_TICKET or raw text pasted by the user.
Boundary markers: Absent. The skill does not use delimiters or instructions to ignore embedded commands within the processed data.
Capability inventory: The agent has access to bash for executing CLI tools.
Sanitization: Absent. There is no evidence of filtering or escaping the fetched ticket content before it is processed for summarization.
[DATA_EXFILTRATION]: The skill accesses and processes sensitive PII, including customer names, email addresses, and private support conversation history from the Gorgias platform. While this is within the scope of its stated purpose, the ingestion of this data into the LLM context represents a data exposure risk if the agent is further manipulated.

ticket-summarize