langfuse-incident-runbook
Pass
Audited by Gen Agent Trust Hub on Sep 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses standard command-line tools such as
curl,grep,docker, andjqto perform diagnostic checks on service status, API connectivity, and local container health. These operations are consistent with the troubleshooting scope defined in the runbook. - [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted data from application logs, container logs, and external API responses during incident triage.
- Ingestion points: Application logs (
SKILL.md), container logs (SKILL.md), and API data (SKILL.md,references/implementation.md). - Boundary markers: Absent in prompt instructions.
- Capability inventory:
curl(SKILL.md,references/implementation.md),dockerCLI (SKILL.md),grep(SKILL.md), and file writing viafs.writeFileSync(references/implementation.md). - Sanitization: Present; instructions in
SKILL.mdexplicitly direct the agent to exclude raw credentials and customer payloads from produced incident records.
Audit Metadata