sentry-create-alert
Warn
Audited by Gen Agent Trust Hub on Feb 21, 2026
Risk Level: MEDIUMCOMMAND_EXECUTION
Full Analysis
- COMMAND_EXECUTION (MEDIUM): The skill relies on executing
curlandpython3commands via the shell, interpolating user-provided variables directly into command strings. - Evidence: Phase 2, 4, and 'Managing Alerts' sections use shell interpolation for
{region},{org},{token}, and{payload}. For example:curl -s "https://{region}.sentry.io/api/0/organizations/{org}/members/". - Risk: Malicious user input containing shell metacharacters (e.g.,
;,|,$(...)) could be used to execute unauthorized commands on the host system if the agent lacks input sanitization logic. - DATA_EXFILTRATION (SAFE): Network activity is restricted to
*.sentry.iodomains for legitimate API interactions. No evidence of data being sent to untrusted third-party servers was found. - CREDENTIALS_UNSAFE (SAFE): The skill correctly uses placeholders (
{token}) and instructs the agent to collect credentials from the user at runtime. No hardcoded API keys or secrets are present in the code. - REMOTE_CODE_EXECUTION (SAFE): While the skill uses
python3 -c, it is used to execute static, hardcoded one-liners for JSON processing of local API responses, rather than fetching and executing remote scripts.
Audit Metadata