alerting-rules
Pass
Audited by Gen Agent Trust Hub on Oct 27, 2026
Risk Level: SAFECREDENTIALS_UNSAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill instructions involve reading and editing the Alertmanager configuration file, which is a high-value target for credential harvesting.
- Evidence: The workflow references path
/opt/homelab/docker/monitoring/alertmanager/alertmanager.yml. - Details: This configuration file typically stores sensitive information including SMTP credentials, Slack/PagerDuty webhook URLs, and ntfy access tokens.
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the monitoring system and interpolates it into configuration rules.
- Ingestion points: Metrics retrieved via
homelab_prometheusQueryand current alerts fromhomelab_alertList(SKILL.md). - Boundary markers: The skill lacks instructions for the agent to treat tool outputs as untrusted data or to use delimiters when processing metric values or labels.
- Capability inventory: The skill has the ability to write to system configuration files and manage monitoring services (SKILL.md).
- Sanitization: There is no evidence of sanitization or escaping of metric labels or values before they are used to generate alerting rules (e.g., in alert annotations).
Audit Metadata