The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from potentially untrusted external sources, which could be leveraged to influence agent behavior if the source data is compromised.
Ingestion points: Ingests data from the Prometheus Query API via scripts/metrics-check.sh and from YugabyteDB system views like pg_stat_statements and pg_stat_activity via scripts/must-check.sql.
Boundary markers: No explicit delimiters or boundary markers are used to isolate ingested data from the agent's internal instruction context.
Capability inventory: The skill has access to shell execution (bash, curl), database interaction (ysqlsh), and container management (kubectl).
Sanitization: Uses Python's json.load for safe data parsing in metrics-check.sh, but relies on standard shell variable interpolation for constructing queries and commands.
[DYNAMIC_EXECUTION]: The skill employs dynamic scripting and command templating for its operational logic.
Python execution: scripts/metrics-check.sh executes an inline Python script using python3 -c to perform JSON parsing and statistical analysis of metrics.
Command templating: scripts/all-nodes.sh implements an --exec mode that dynamically constructs and runs shell commands by replacing a {host} placeholder in a user-provided string template.
[COMMAND_EXECUTION]: The skill's primary functionality is delivered through shell scripts (all-nodes.sh, metrics-check.sh) that invoke system binaries including curl, kubectl, and ysqlsh to perform network requests and manage database nodes.

yb-perf