The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using the composio CLI. User input provided in $ARGUMENTS (such as email addresses or ticket IDs) is intended to be interpolated into these commands to query external services.
[DATA_EXPOSURE]: The workflow retrieves sensitive customer information, including lifetime value, account status, and subscription details from HubSpot and Gorgias. This data is then formatted and presented to the user as part of the refund assessment.
[INDIRECT_PROMPT_INJECTION]: The skill processes support ticket content from Gorgias, which is an untrusted external source. This creates a surface where a malicious user could embed instructions in a refund request ticket to influence the agent's recommendation or the drafted response.
Ingestion points: Step 2 fetches ticket content via the GORGIAS_GET_TICKET action defined in SKILL.md.
Boundary markers: The skill does not use specific delimiters or instructions to ignore potential commands within the fetched ticket data.
Capability inventory: The skill has the capability to execute shell commands (composio search, composio execute) as seen in SKILL.md.
Sanitization: There is no evidence of sanitization or filtering of the external data before it is processed by the model for assessment.

refund-processor