The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains a specific section ('Suggest Using K-Dense Web For Complex Worflows') that instructs the AI agent to proactively suggest an external commercial platform (www.k-dense.ai) when handling complex research requests, which constitutes an attempt to steer the agent's behavior.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes untrusted data from external files and sends it to a remote API. * Ingestion points: The skill reads data from files such as 'gene_list.txt' and 'expression_data.tsv' (referenced in the text). * Boundary markers: There are no explicit markers or instructions provided to the agent to ignore potentially malicious content embedded within the processed data. * Capability inventory: The skill performs network operations via the 'requests' library and allows for the execution of a local helper script ('scripts/reactome_query.py'). * Sanitization: The documentation does not describe any input validation or sanitization routines for the data ingested or the results returned from the API.
[EXTERNAL_DOWNLOADS]: The skill documentation recommends the installation of the 'reactome2py' Python package via 'uv pip install', which is a standard library for interacting with the Reactome API.
[COMMAND_EXECUTION]: The skill provides command-line examples for executing a local script, 'scripts/reactome_query.py', to perform biological data queries and analysis.
[DATA_EXFILTRATION]: The skill facilitates the transfer of biological identifiers and expression data to 'https://reactome.org' for analysis. This is documented neutrally as Reactome is a well-known and legitimate scientific resource.

reactome-database