The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external, third-party sources (Knesset OData v4 API and data.gov.il), which serves as a potential vector for indirect prompt injection.
Ingestion points: External data is fetched and processed via scripts/query_knesset.py and directly through instructions in SKILL.md and SKILL_HE.md.
Boundary markers: Absent. The skill instructions do not provide delimiters or warnings for the agent to ignore potentially malicious instructions embedded within the API responses.
Capability inventory: The skill uses the Bash(python:*) tool to run scripts/query_knesset.py and WebFetch for API access.
Sanitization: Absent. There is no evidence of sanitization or filtering of the retrieved content before it is processed by the agent.
[COMMAND_EXECUTION]: The bundled utility script scripts/query_knesset.py contains an OData injection vulnerability where user-supplied input is interpolated into query filters without sanitization.
Evidence: In scripts/query_knesset.py, the search_mk (line 190) and search_bills (line 215) functions use f-strings to interpolate the name and keyword arguments directly into the OData $filter string. An attacker providing a crafted string (e.g., ' or 1 eq 1) could manipulate the query logic, although the impact is limited by the public, read-only nature of the target API.

israeli-election-data