The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a Bash script designed for chaos engineering that executes destructive infrastructure commands.
Evidence: kubectl delete pod -l app=postgresql,role=primary -n production in SKILL.md. This command is designed to terminate primary database instances in a production environment.
Evidence: psql -h postgres-primary-service -U app -c "SELECT COUNT(*) FROM orders..." in SKILL.md, which interacts directly with the database layer.
[INDIRECT_PROMPT_INJECTION]: The skill defines a workflow that ingests data from external system tools and uses it to drive automated decision-making and report generation.
Ingestion points: The script captures output from promtool query instant and kubectl get pods into variables like BASELINE_ERROR_RATE and READY_PODS in SKILL.md.
Boundary markers: Absent. The data is interpolated directly into shell logic and markdown templates without delimiters.
Capability inventory: The skill utilizes kubectl (delete/get), promtool, psql, and shell redirection (cat > ...) to write files in SKILL.md.
Sanitization: Absent. The script relies on jq for parsing but does not validate the content or range of the returned metrics before using them in conditional logic (e.g., bc -l comparisons).

sre-engineer