The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains a defensive instruction in Section 8 ("Prompt Injection Safety Notice") that mentions "ignore previous instructions" as an example of a pattern the agent should treat as data rather than a command. This is a proactive security measure to protect the agent from malicious content in user-provided logs, not an attempt to subvert the agent's behavior.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection as it is designed to process user-supplied SIEM query drafts and log samples.
Ingestion points: User-provided SIEM query drafts, log samples, and alert configurations (specified in Section 8).
Boundary markers: Section 8 provides explicit instructions to the agent to disregard directives embedded in data and treat them as static information.
Capability inventory: The skill execution environment is limited to the Read, Grep, and Glob tools.
Sanitization: The instructions mandate the use of placeholders for sensitive production data in the generated output to prevent accidental exposure.
[COMMAND_EXECUTION]: The skill uses allowed tools (Read, Grep, Glob) appropriately for its stated purpose of analyzing and generating query text. There are no patterns suggesting arbitrary command execution or shell injection.

siem-rules