The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from manuscript files and BibTeX databases, which could allow maliciously crafted content to influence the agent's behavior during the analysis phases.
Ingestion points: Manuscript files (markdown, docx, txt) and references.bib files.
Boundary markers: Absent; the skill does not define delimiters or provide instructions to ignore instructions embedded in the processed data.
Capability inventory: Execution of shell commands (git, grep, pandoc), file system operations (read/write), and Python-based regex parsing.
Sanitization: Absent; the skill processes input data directly without prior validation or escaping.
[COMMAND_EXECUTION]: The workflow requires the agent to execute shell commands like git, grep, and pandoc using arguments (such as citation keys or filenames) derived from external user-provided files. This presents a surface for potential command injection if the agent does not correctly handle or quote these arguments.
[EXTERNAL_DOWNLOADS]: The skill recommends downloading CSL style files from the official Citation Style Language GitHub repository. This is a well-known and standard service for academic bibliography formatting.

bibliography-builder