The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from upstream research processes (such as experiment plans and trajectory logs), which introduces a potential vulnerability surface for indirect prompt injection.
Ingestion points: The skill pulls summary artifacts and experimental results from skills like paper-planning and experiment-pipeline as described in SKILL.md.
Boundary markers: There are no explicit instructions or delimiters defined to separate ingested data from the skill's own instructions or to ignore embedded commands.
Capability inventory: The skill has access to file-modifying capabilities through write_file, edit_file, and read_file.
Sanitization: No explicit sanitization, filtering, or validation steps are included for the data retrieved from external skill artifacts before they are used in the drafting process.

paper-writing