The Agent Skills Directory

[SAFE]: The skill's instructions, tool definitions, and workflows are entirely consistent with its stated purpose of PagerDuty automation. No malicious behaviors such as credential theft, unauthorized file access, or persistence mechanisms were detected.
[PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect prompt injection. It ingests untrusted data from external sources, including PagerDuty incident notes and alert details, through tools like PAGERDUTY_FETCH_INCIDENT_LIST, PAGERDUTY_RETRIEVE_INCIDENT_BY_INCIDENT_ID, and PAGERDUTY_GET_ALERTS_BY_INCIDENT_ID (SKILL.md). The instructions do not define boundary markers or provide explicit warnings to the agent to ignore instructions embedded within this data. Given that the skill also grants write capabilities, such as updating incident statuses (PAGERDUTY_UPDATE_INCIDENT_BY_ID) and managing schedules (PAGERDUTY_UPDATE_SCHEDULE_BY_ID), an attacker could potentially influence agent actions through malicious content in a PagerDuty incident note.

pagerduty-automation