The Agent Skills Directory

[COMMAND_EXECUTION]: The SKILL.md file contains instructions for the agent to use bash along with python3 or jq to process large tool outputs that have been saved to local files. This practice is specifically intended to manage large datasets that exceed the standard context window limits.
[PROMPT_INJECTION]: The skill involves processing untrusted external data (logs, traces, error messages), creating an indirect prompt injection surface.
Ingestion points: Observability data is ingested through several tools, including query-logs, query-traces, and query-sessions as described in SKILL.md and referenced in companion files like logs.md and traces.md.
Boundary markers: There are no specific instructions or delimiters mentioned to isolate ingested data from the agent's core instructions.
Capability inventory: The skill specifically grants the agent the ability to execute shell commands (bash, python3, jq) to manipulate files containing tool results.
Sanitization: The skill does not describe or require sanitization steps for data retrieved from the observability backend before it is processed by the agent.

investigate