The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from a remote Memos instance which may contain untrusted user content.\n
Ingestion points: Data is fetched via scripts/memos.sh using the list, get, and comments commands.\n
Boundary markers: There are no explicit delimiters or instructions provided to the agent to distinguish fetched memo content from operational instructions.\n
Capability inventory: The skill is capable of network communication via curl and data manipulation using python3 and jq.\n
Sanitization: The script does not perform textual sanitization on the fetched content before returning it to the agent.\n- [COMMAND_EXECUTION]: The skill performs shell command execution to interface with the REST API.\n
It uses curl for all network requests to the Memos server.\n
It uses jq to parse and build JSON objects safely.\n- [EXTERNAL_DOWNLOADS]: The skill communicates with external servers to manage memos.\n
Requests are sent to the user-defined MEMOS_URL using authentication tokens from MEMOS_ACCESS_TOKEN.\n- [DYNAMIC_EXECUTION]: The skill executes short Python snippets to handle data formatting.\n
python3 -c is used for URL-encoding and JSON array construction.\n
These snippets use safe argument passing via sys.argv instead of string interpolation.

memos