The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read and summarize structured application logs (NDJSON) from the .evlog/logs/ directory. These logs contain data from HTTP requests and error messages which are influenced by untrusted external users.
Ingestion points: The agent reads content from .jsonl files located in the project's .evlog/logs/ directory.
Boundary markers: The instructions do not specify any delimiters or warnings to ignore instructions that might be embedded within the log data fields (e.g., in error.message or error.data.why).
Capability inventory: The agent has the capability to read files, provide explanations to the user, and suggest the execution of shell commands (npx).
Sanitization: There is no mention of sanitizing or escaping the log content before it is processed or presented to the user.
[COMMAND_EXECUTION]: The skill suggests running npx @evlog/cli init and npx @evlog/cli doctor to set up or diagnose log collection. Executing CLI tools involves the execution of shell commands.
[REMOTE_CODE_EXECUTION]: The use of npx for @evlog/cli involves downloading and executing code from the npm registry. While these are resources provided by the vendor, they represent an external code execution pattern.

analyze-logs