The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and forensically analyze text extracted from untrusted PDF manuscripts provided by the user. Maliciously crafted content within these documents could attempt to influence the agent's behavior, subagent reporting, or the master editor's final recommendations.
Ingestion points: Text extraction from user-provided PDFs in Step 0 (SKILL.md) using pdftotext or Python libraries.
Boundary markers: Absent; the skill does not implement specific delimiters or 'ignore' instructions for the extracted content before it is processed by subagents.
Capability inventory: The skill has access to Bash, Write, Edit, and Read tools, providing a significant surface for impact if an injection occurs.
Sanitization: None; the agent is instructed to read the full extracted text end-to-end without filtering.
[COMMAND_EXECUTION]: The skill explicitly instructs the agent to use the Bash tool to perform text extraction, directory management, and report generation. The use of a shell environment to process data derived from untrusted external sources increases the potential impact of an injection attack, even though the commands (e.g., pdftotext) are standard utilities.

academic-proofreader