The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process content from various untrusted source files, creating a surface for indirect prompt injection.
Ingestion points: The skill reads extensively from transcript.md, .audit/*.md files, and session records in .records/ to make curriculum decisions and re-plan units.
Boundary markers: The instructions do not specify the use of clear delimiters or safety prompts when the agent interpolates content from these potentially untrusted sources into its own reasoning or generated artifacts.
Capability inventory: The skill possesses significant file-writing capabilities across the workspace, including the ability to write a HANDOVER.md file to a path specified by the user, which could be exploited to target sensitive file locations if user input is not validated.
Sanitization: Although the skill specifies a 'translation' process for roadmap rendering, it requires the agent to quote failed audit findings in full when generating remediation steps in unit.md, which could lead to the direct interpolation of malicious instructions.

academic-advisor