The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to execute a local Python script (search_ai_papers.py) to perform novelty verification against academic databases including arXiv, Semantic Scholar, and OpenAlex. This relies on the existence of a shared research skill in the local environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from research papers, external codebases, and user-provided ideas, and interpolates this content into prompts for the evaluation agent.
Ingestion points: Data is read from Ideation/references/papers/ (LaTeX files), Experiment/code_references/ (source code and READMEs), and selected_idea.txt.
Boundary markers: Evaluation templates use Markdown headers to separate sections, but they do not employ explicit boundary delimiters or instructions for the agent to disregard instructions embedded in the ingested data.
Capability inventory: The skill has the ability to read and write files within the project workspace and execute shell commands via a local script.
Sanitization: There is no evidence of sanitization, filtering, or escaping of the ingested content before it is provided to the Eval Agent in the evidence_block.

inno-idea-eval