The Agent Skills Directory

[COMMAND_EXECUTION]: The skill wraps the logcli binary using shell scripts to interact with Loki instances. It handles arguments securely and restricts execution to a predefined set of internal command scripts.
[EXTERNAL_DOWNLOADS]: The skill is configured to pull the grafana/logcli Docker image from Docker Hub if a local installation is not found. This targets an official image from a well-known service provider and is used as a runtime environment.
[PROMPT_INJECTION]: The skill processes untrusted log data from external sources, which presents a surface for indirect prompt injection. 1. Ingestion points: Logs fetched via scripts/commands/logs/query.sh. 2. Boundary markers: The skill does not use specific delimiters or instructions to isolate log content. 3. Capability inventory: Read-only log querying and label inspection. 4. Sanitization: External log content is output directly without escaping or filtering.
[DATA_EXFILTRATION]: A test file (tests/test_env_resolution.sh) contains a hardcoded absolute path to a developer's local directory (/Users/Dmytro/Projects/vinitu/logcli-skill/.env). While this is a poor development practice, it does not constitute an exfiltration attempt.

logcli