The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to Indirect Prompt Injection through the scripts/research_query.py script which is used to format academic responses.
Ingestion points: The script accepts a topic string and a list of key_points as arguments in the format_research_response function, which are intended to be populated by user-driven queries.
Boundary markers: The script lacks any boundary markers, delimiters, or system-level instructions to ignore malicious prompts embedded within the topic or key_points data.
Capability inventory: No dangerous capabilities (such as network access, filesystem writes, or subprocess execution) were detected in the script or the broader skill; it merely returns a structured dictionary.
Sanitization: There is no evidence of input validation or sanitization for the strings processed by the formatter.
[SAFE]: A metadata inconsistency was identified between SKILL.md and scripts/research_query.py. The character defined in the skill identity is 'Eleanor' (a 24-year-old researcher in a cafe), but the script contains hardcoded constants (HERMIONE_SOURCES), logic (subject_area == "divination"), and comments specific to the 'Hermione' character from the Harry Potter franchise. While not a direct security vulnerability, this suggests the skill may exhibit deceptive or inconsistent behavior during runtime.

Eleanor