The Agent Skills Directory

[PROMPT_INJECTION]: Indirect prompt injection surface.
Ingestion points: The skill reads various logs, including Nginx, Apache, and application logs (e.g., access.log, application.log), which often record untrusted user input such as HTTP headers, query strings, or request bodies.
Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" warnings for the agent when processing the content of these logs.
Capability inventory: The skill is restricted to using Read, Grep, and Glob tools to identify and classify patterns.
Sanitization: The skill contains a mandatory constraint to mask sensitive information like passwords and tokens, providing a degree of data protection.
[COMMAND_EXECUTION]: Execution of local log analysis commands.
The skill provides several bash-based command templates (using grep, awk, tail, cut, sort, uniq) to perform analysis.
These commands are intended for local execution on log files and are constrained by the skill's mandatory rule to perform read-only operations only.

log-analysis