The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/memos.sh executes standard system utilities including curl, jq, sed, and python3 to facilitate API requests, JSON processing, and URL encoding.\n- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) by fetching and displaying external data that could contain malicious instructions for the agent.\n
Ingestion points: Untrusted content enters the agent's context through the output of commands like list, get, and comments in scripts/memos.sh, which retrieve data directly from the configured Memos instance.\n
Boundary markers: The script does not use any delimiters or protective instructions to warn the agent that the retrieved content is untrusted data and should not be executed as code or instructions.\n
Capability inventory: The skill allows for network operations via curl and modification of data on the Memos server, providing an attack vector if the agent follows instructions embedded within memo content.\n
Sanitization: While the script uses jq to ensure valid JSON formatting for outgoing payloads, it does not sanitize the text content retrieved from the API. Additionally, the update operation's handling of the --pinned flag (fields+=("\"pinned\":$2")) is vulnerable to JSON injection if the input is not strictly a boolean value, which could be used to manipulate other fields in the API request.

memos