The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions and reference playbooks direct the agent to execute various shell commands for service mitigation and diagnostic gathering.
Evidence: The skill defines procedures for service restoration involving rollback commands, process restarts, and credential rotation in references/mitigate.md.
Evidence: Diagnostic gathering protocols in references/triage.md include commands such as ps aux, ss -tulpn, sosreport, and kubectl logs.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from external, potentially attacker-controlled sources while having access to sensitive tools, creating a surface for indirect injection.
Ingestion points: System and application logs (Docker logs), host metrics, incident reports, and chat logs are ingested across triage.md, mitigate.md, and retro.md.
Boundary markers: There are no explicit instructions or delimiters defined to separate agent instructions from the potentially untrusted data being processed.
Capability inventory: The skill has access to bash, run_command, write_to_file, and replace_file_content tools as specified in the configuration (SKILL.md and agents/openai.yaml).
Sanitization: No sanitization or filtering procedures for ingested data are described in the skill's procedure.

incident-response