triage-frontend-issues
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external Sentry issue reports, such as issue titles, culprits, and stack frames.\n
- Ingestion points: Data is fetched from the Sentry platform using
mcp__sentry__get_sentry_resourcein the workflow defined inSKILL.md.\n - Boundary markers: The skill does not implement explicit boundary delimiters for the ingested data; however, it follows a strict classification logic based on a local reference file (
references/archive-criteria.md).\n - Capability inventory: The skill's capabilities are limited to searching and updating issue statuses to 'ignored' via
mcp__sentry__update_issue. It lacks the ability to execute arbitrary code, write files, or perform non-whitelisted network operations.\n - Sanitization: The skill mitigates potential injection risks by requiring a mandatory human review step. The agent must present a 'Triage plan' to the user and wait for explicit approval ('apply') before performing any updates, ensuring the user remains in control of all state changes.
Audit Metadata