The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to retrieve a sensitive credential (the Redmine API token) from the system's environment variables ($env:REDMINE_API_TOKEN or echo $REDMINE_API_TOKEN) and transmit it to an external domain (redmine-skgd3-local.housei-inc.com) via URL parameters or headers. While this is the intended primary purpose of the skill, it involves the handling and network transmission of sensitive authentication data to a non-whitelisted remote endpoint.
[INDIRECT_PROMPT_INJECTION]: The skill processes ticket data from the external Redmine instance, including the description and subject fields. This content is untrusted and could contain malicious instructions aimed at influencing the agent's logic or bypassing its constraints.
Ingestion points: Data returned from Redmine API endpoints (/issues.json, /issues/{id}.json) referenced in SKILL.md.
Boundary markers: None identified. The instructions do not specify using delimiters or warnings to ignore instructions within the retrieved ticket data.
Capability inventory: The skill uses terminal access to read environment variables and utilizes the fetch_webpage tool for network communication.
Sanitization: No sanitization or escaping of the ingested ticket content is mentioned before it is processed by the agent.

redmine-search