The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to utilize standard shell utilities, including ls, head, cut, sort, wc, and grep, to inspect, filter, and calculate statistics from local forecasting results.
[PROMPT_INJECTION]: The skill is subject to indirect prompt injection risks because it ingests and interprets data from external files.
Ingestion points: Reads metrics from results_*.csv files located in the nixtla_baseline_m4/ directory (SKILL.md).
Boundary markers: The instructions do not define clear delimiters or provide warnings to the agent to ignore potentially malicious instructions embedded within the CSV data.
Capability inventory: The skill utilizes Read, Grep, and Bash tools, allowing it to read files and execute shell commands based on findings in the data (SKILL.md).
Sanitization: There is no evidence of content validation, escaping, or filtering for the external CSV data before it is processed by the agent.

nixtla-baseline-review