The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's bash scripts and inline Python snippets (using python3 -c) unsafely interpolate shell variables like ${TRACE_ID}, ${ENV_KEY}, and ${KEYWORD} directly into executable code. A malicious user could provide input containing shell metacharacters or Python code to execute arbitrary commands within the agent's environment.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it fetches and displays raw log data from external Grafana Loki instances (ingestion point in SKILL.md). Malicious instructions embedded in the logs could influence the agent's actions, as there are no sanitization steps or strong boundary markers to separate untrusted data from instructions.
[DATA_EXFILTRATION]: The skill accesses a local configuration file (environments.json) containing sensitive authentication tokens and transmits them to external endpoints. The use of unvalidated environment keys to access these tokens increases the risk of unauthorized credential exposure or exfiltration if the environment is manipulated.
[EXTERNAL_DOWNLOADS]: The skill performs network requests to multiple external Grafana instances (e.g., test13.xnzn.net, monitor-test.xnzn.net) to retrieve log data. These requests are triggered by user input and carry authentication tokens.

loki-log-query