triage-frontend-issues
Pass
Audited by Gen Agent Trust Hub on Jun 22, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXFILTRATION]: All data access and mutations are restricted to the Sentry organization via authenticated MCP tools. The skill does not transmit data to external or untrusted third-party domains. Operations are limited to the specific project slug
javascript. - [PROMPT_INJECTION]: The skill ingests untrusted data from Sentry issue metadata (such as error titles and stack traces) which represents an indirect prompt injection surface. However, this risk is mitigated by hard-coded constraints that prevent any action other than archiving and a requirement for explicit human approval of a proposed triage plan before execution.
- [COMMAND_EXECUTION]: The skill relies on structured MCP tool calls (
search_issues,get_sentry_resource,update_issue) rather than arbitrary shell commands. It does not use dynamic code execution or subprocess spawning. - [SAFE]: The skill implements robust 'Hard Rules' that prevent common pitfalls such as accidental deletion, unauthorized assignment changes, or broad project access. The use of the
untilEscalatingmode ensures that archived issues will reappear if conditions worsen, providing a safety net against incorrect triage decisions.
Audit Metadata