The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests research paper data from the Semantic Scholar API and interpolates it into prompts for novelty assessment. While the risk is present whenever processing external data, the skill uses standard mitigation practices.
Ingestion points: The novelty_check.py script retrieves paper titles and abstracts from Semantic Scholar, which are then passed into prompt templates in references/ideation-prompts.md using the {last_query_results} variable.
Boundary markers: The prompt templates utilize triple quotes (""") and XML-style tags (<experiment.py>) to delimit external content from instructions.
Capability inventory: The skill executes a local Python script which performs network operations to external APIs to retrieve data.
Sanitization: No explicit sanitization or filtering of the retrieved paper abstracts is performed before they are interpolated into the LLM prompts.
[EXTERNAL_DOWNLOADS]: The novelty_check.py script makes network requests to the Semantic Scholar official API (api.semanticscholar.org) to retrieve research literature. This is a well-known service and the operation is central to the skill's stated purpose of checking research novelty.

idea-generation