The Agent Skills Directory

[PROMPT_INJECTION]: The skill is designed to ingest and analyze external Docker logs and error messages, which constitutes an indirect prompt injection surface.
Ingestion points: Processes output from commands like docker compose logs and terminal error strings.
Boundary markers: Absent. The instructions do not provide delimiters or warnings to the agent to distinguish between log content and operational instructions.
Capability inventory: The skill utilizes shell commands for Docker management (docker compose), network inspection (lsof, ss), and service control (systemctl).
Sanitization: No sanitization or validation of the ingested log content is performed before analysis.
[COMMAND_EXECUTION]: The diagnostic guides suggest the use of administrative commands such as sudo systemctl restart docker and sudo chown. While these are legitimate for Docker troubleshooting, they involve privilege escalation and direct interaction with host system services.
[PROMPT_INJECTION]: There is a discrepancy in the author metadata; the YAML frontmatter identifies the author as 'OpenAEC-Foundation', whereas the skill context identifies 'Impertio-Studio'. This inconsistency in identifying the source can be considered a minor metadata poisoning issue.

docker-errors-compose