The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from an external Cognee server and session memory, creating a vulnerability surface where malicious instructions embedded in the memory could influence the agent's behavior.
Ingestion points: Data is retrieved via curl and the cognee-search.sh script from POST /api/v1/recall and local CLI output in SKILL.md.
Boundary markers: No explicit delimiters or instructions are provided to the agent to disregard potential instructions within the retrieved memory content.
Capability inventory: The skill utilizes shell command execution capabilities through curl, cognee-cli, and local script execution.
Sanitization: There is no evidence of sanitization or filtering applied to the data returned from the Cognee server before it is processed by the agent.
[COMMAND_EXECUTION]: The skill relies on executing shell commands (curl, cognee-cli, and a provided shell script) to perform its core search functionality.
Evidence: The instructions demonstrate the use of ${CLAUDE_PLUGIN_ROOT}/scripts/cognee-search.sh and direct curl commands to interact with the Cognee API.
Context: While these commands are intended for searching memory, they represent powerful capabilities that could be misused if the agent's context is manipulated via indirect injection.

cognee-search