The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill relies on the agent constructing and executing Javascript programs at runtime. The instructions guide the agent to interpolate user-provided inputs (like names and context) into predefined script templates which are then run through the arcmira_execute_read tool.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted external data which could contain malicious instructions.
Ingestion points: External data is pulled into the agent context via arcmira.search results and arcmira.transcript content in SKILL.md.
Boundary markers: The skill lacks explicit boundary markers or instructions to the model to ignore instructions embedded within the research data.
Capability inventory: The skill utilizes arcmira_execute_read for data fetching and logic execution, and references arcmira_execute_write for other operations.
Sanitization: While the script performs basic string slicing for display purposes, it does not implement sanitization or escaping to prevent potential prompt injection contained within video transcripts or search results.

person-research