The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from Canvas assignments and submissions. Malicious instructions embedded in these fields by students could potentially manipulate the agent's behavior during report generation or when drafting follow-up messages.\n
Ingestion points: Data enters the agent context through the list_assignments and list_submissions tools as specified in SKILL.md.\n
Boundary markers: The skill does not define specific delimiters or "ignore embedded instructions" warnings to separate untrusted Canvas data from the agent's system instructions.\n
Capability inventory: The agent possesses write capabilities through tools like send_conversation and send_peer_review_inbox_messages, which could be triggered by injected instructions.\n
Sanitization: There is no evidence of content sanitization or validation for the data retrieved from the LMS before it is interpolated into prompts.

canvas-morning-check