The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses Ansible to execute a variety of system and Docker commands on remote hosts to gather diagnostic data.
Evidence: Instructions in SKILL.md and references/standards.md specify the execution of docker ps, docker logs, docker stats, docker service ps, df -h, df -i, and systemctl --failed across hosts defined in the inventory.
Mitigation: The skill emphasizes a strictly "read-only" posture, explicitly forbidding destructive commands (e.g., docker system prune, reboot, apt) and requiring host confirmation via ./deploy check-hosts before execution.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from remote server logs and system outputs, which are then interpolated into Markdown reports.
Ingestion points: System command outputs and Docker container logs (docker logs --tail 200) from remote hosts are read and included in the report files generated in SKILL.md and references/report-template.md.
Boundary markers: The skill utilizes Markdown code blocks (e.g., ```text) to encapsulate external data in the report template, providing structural delimitation but not preventing natural language interpretation by the agent.
Capability inventory: The agent has the capability to execute shell commands via Ansible on remote hosts and write files to the local server-reports/ directory.
Sanitization: The skill provides clear instructions to mask secrets, tokens, and sensitive variables using no_log: true and post-generation grep checks. However, there are no specific instructions to sanitize or escape Markdown control characters or potential instruction patterns embedded within the remote logs themselves.

specsfy-specialist-server-report