The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill exposes a vulnerability surface for indirect prompt injection (Category 8) by instructing the agent to process untrusted data from production logs and traces while maintaining access to infrastructure tools.
Ingestion points: The agent is guided to ingest and search through production logs, distributed traces (trace_id), and alert descriptions in SKILL.md and references/common-failure-modes.md.
Boundary markers: No explicit delimiters or instructions are provided to isolate untrusted data from the agent's command logic.
Capability inventory: The skill utilizes powerful command-line interfaces including kubectl, docker, aws-cli, git, and redis-cli for diagnostics.
Sanitization: There is no logic or instruction for sanitizing or escaping content from logs before the agent processes it.
Command Execution (SAFE): The skill provides numerous shell command templates for Kubernetes, Docker, and system diagnostics (e.g., top, df, netstat). These commands are standard for the skill's stated purpose of incident response and do not perform unauthorized or destructive actions.

incident-response