The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external research papers, including their "scientific notes" and output files. This creates a surface for indirect prompt injection where malicious instructions embedded in a paper could influence the agent's behavior.
Ingestion points: The agent resolves and reads the "active paper", its "scientific notes", and associated documentation (SKILL.md).
Capability inventory: The agent is instructed to re-run procedures and process file contents (references/report-gate.md).
Boundary markers: There are no explicit instructions to the agent to disregard instructions found within the research papers.
Sanitization: The skill does not describe any sanitization or validation of the contents of the paper or its notes.
[COMMAND_EXECUTION]: The verification process requires the agent to "Re-run the paper's recorded procedure on its recorded inputs" (references/report-gate.md). This involves executing code found within the research project's directory, which could lead to arbitrary command execution if the project contains malicious scripts.
[DYNAMIC_EXECUTION]: The skill facilitates the execution of scripts or build procedures defined in the research paper to "regenerate the headline number". This dynamic execution of local artifacts is a core function but represents a security risk if the paper being analyzed is from an untrusted or compromised source.

rk-report