The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to invoke the tectonic and pandoc command-line utilities to compile LaTeX source files into PDF documents during the final stages of the pipeline. This is a legitimate functional requirement for the skill's stated goal of manuscript preparation.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests and processes untrusted content from user-provided paper drafts and reviewer feedback.
Ingestion points: Untrusted data enters the context in agents/pipeline_orchestrator_agent.md and agents/integrity_verification_agent.md during the drafting, integrity verification, and revision phases.
Boundary markers: The instructions do not define specific delimiters or instructions to mitigate the risk of the agent obeying malicious commands embedded within the processed papers.
Capability inventory: The skill maintains access to capabilities such as WebSearch and shell command execution for document compilation.
Sanitization: There are no documented procedures for sanitizing or escaping the content of external documents before they are interpolated into agent prompts.

academic-pipeline