The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it fetches and processes content from external conference websites (CFPs) via WebFetch to determine submission constraints. Malicious content on these third-party sites could attempt to subvert the agent's instructions. * Ingestion points: The official CFP is fetched during Phase 0 to extract page limits and policies from SKILL.md. * Boundary markers: There are no explicit delimiters or instructions to ignore embedded commands within the fetched CFP content. * Capability inventory: The agent has access to Bash (used for latexmk, pdftotext, grep), Write, and Edit tools across the workflow. * Sanitization: No sanitization or validation of the external CFP content is performed before it influences the workflow.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute standard document processing commands including latexmk, pdftotext, kpsewhich, and grep. It also uses mkdir and ln for file system organization and curl for a POST request to a local synchronization API on localhost:8765 in Phase 5. These actions are functionally relevant to the stated purpose of paper management and local environment integration.
[EXTERNAL_DOWNLOADS]: The skill fetches external data from Call for Papers (CFP) websites using the WebFetch tool. This is a necessary step for verifying venue compliance but introduces a dependency on untrusted external sources that could provide malicious input to the agent context.

venue-fork