The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because its core workflow requires reading and processing untrusted external data.
Ingestion points: The skill explicitly instructs the agent to read "project context", "progress logs", "current focus", and "recent work" from the local filesystem.
Boundary markers: There are no instructions provided to wrap external content in delimiters or to ignore potential instructions embedded within those files.
Capability inventory: The agent has permissions for file operations (Read, Write, Edit) and system command execution via Bash for LaTeX compilation.
Sanitization: No sanitization or validation logic is defined to filter malicious content from the ingested project data before it is processed.
[COMMAND_EXECUTION]: The skill utilizes Bash to execute LaTeX compilation tools (latexmk, xelatex, pdflatex). While necessary for the skill's primary purpose of generating PDF presentations, executing these binaries via shell can be a security risk if the LaTeX environment is not properly sandboxed, particularly if features like shell-escape are enabled.

project-deck