The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external transcriptomics data files which could be used to influence the agent's behavior if the data is untrusted.
Ingestion points: Files such as expression_data.tsv and tf_list.txt are read into the agent's context in scripts/basic_grn_inference.py and referenced in SKILL.md.
Boundary markers: There are no protective delimiters or instructions to ignore potential commands embedded in the data.
Capability inventory: The skill allows writing analysis results to the local filesystem and establishing network connections to Dask schedulers.
Sanitization: Input data is not validated or sanitized for potentially malicious strings or structural anomalies.
[DATA_EXFILTRATION]: The skill documents and enables network connections to remote Dask schedulers via TCP addresses (e.g., tcp://scheduler:8786). This constitutes a network operation to non-whitelisted endpoints, which represents a potential channel for data transmission to external infrastructure.

arboreto