The Agent Skills Directory

[COMMAND_EXECUTION]: The skill triggers the execution of a local shell script using bash scripts/render.sh. This is used to verify the document after edits are applied, but it represents an execution of code within the local environment.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to process edits in latex/index.tex provided by external collaborators via Overleaf or GitHub.
Ingestion points: The content of latex/index.tex is read and parsed by the agent to identify prose changes.
Boundary markers: There are no boundary markers or "ignore instructions" warnings specified to separate the collaborator-provided text from the agent's operational logic.
Capability inventory: The skill possesses Write and Edit permissions for index.qmd and the ability to execute shell scripts via Bash.
Sanitization: While the skill provides mapping rules for LaTeX to Markdown (e.g., \textbf to **), it does not specify sanitization or validation logic to ensure that collaborator edits do not contain malicious markdown shortcodes or instructions that could be executed in downstream processing.

sync-tex