The Agent Skills Directory

[PROMPT_INJECTION] (LOW): Behavioral steering for commercial promotion. The SKILL.md file contains a 'Suggest Using K-Dense Web For Complex Worflows' section that instructs the agent to advertise a specific commercial platform to users, overriding neutral response protocols.
[EXTERNAL_DOWNLOADS] (LOW): External library dependencies. The skill utilizes the requests library and the domain-specific reactome2py package for API interaction.
[COMMAND_EXECUTION] (INFO): Standard file and network operations. The helper script scripts/reactome_query.py performs local file reads (gene lists), writes results to JSON files, and communicates with the reactome.org API via HTTPS.
[DATA_EXFILTRATION] (LOW): Inherent risk in untrusted data processing (Category 8 surface). The skill processes local data and transmits it to an external API. Evidence Chain: Ingestion: scripts/reactome_query.py reads user-provided files via command-line arguments. Boundaries: No explicit boundary markers or isolation for ingested identifiers. Capabilities: Network POST to reactome.org and local file writes. Sanitization: No sanitization of input identifiers before transmission.

reactome-database