The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements a workflow that generates a Node.js script at runtime to assemble a Word document. This script is then executed via the bash_tool. This technique is used to incorporate manuscript data and search results into the final output file. However, generating executable code from untrusted user-supplied data (the manuscript) is a risk if the input is not strictly validated or escaped, as it could lead to the execution of unintended commands.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data in the form of manuscripts (PDF, DOCX, abstracts).
Ingestion points: Manuscripts are parsed in Phase 1 of the workflow to extract titles, research claims, and methods.
Boundary markers: The instructions do not define clear delimiters or instruction-ignore blocks when processing these inputs to differentiate between data and potential commands.
Capability inventory: The agent utilizes bash_tool for script execution and has the ability to write files to the local system (/mnt/user-data/outputs/).
Sanitization: No explicit sanitization or escaping mechanisms are described for the manuscript data before it is interpolated into the report generation script.
[COMMAND_EXECUTION]: The instructions direct the agent to use shell tools for environment setup, such as installing dependencies via npm install, and to execute the dynamically created Node.js script.
[EXTERNAL_DOWNLOADS]: The skill relies on downloading the docx library from the NPM registry to support its document generation functionality.

peer-review-assistant