The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes and displays data from previous agent sessions, such as titles and observations, which could contain malicious instructions designed to manipulate the agent during the recap process.
Ingestion points: Data retrieved from the memory_sessions and memory_recall tools as defined in SKILL.md.
Boundary markers: The recap output uses Markdown headers and lists, but no specific delimiters or instructions are provided to the agent to disregard commands within the historical data.
Capability inventory: The skill uses platform-provided memory tools to read session history.
Sanitization: There is no evidence of sanitization or escaping of session content before it is presented to the user.
[PROMPT_INJECTION]: The $ARGUMENTS variable is directly interpolated into the prompt instructions. A malicious user could provide input designed to override the agent's system instructions instead of a standard time window.

recap