The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill requires the agent to write a Python script (ingest.py) at runtime to perform data conversion and parsing. It specifically directs the agent to 'write a custom parser' when encountering unknown data formats, which involves generating and executing code based on the structure of arbitrary, potentially untrusted input.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external, untrusted sources such as iMessage logs, email exports (.mbox, .eml), and Twitter archives. These sources could contain malicious instructions or specially crafted content intended to influence the AI's behavior during the wiki compilation and synthesis phases.
Ingestion points: Untrusted content is read from the data/ directory, including email bodies, message histories, and social media exports.
Boundary markers: There are no instructions provided to use boundary markers or to ignore instructions embedded within the source data during the 'absorb' or 'ingest' phases.
Capability inventory: The agent has the authority to write to the filesystem, create new directories, and generate/execute Python scripts.
Sanitization: No sanitization or escaping mechanisms are specified for the untrusted content before it is processed or integrated into the markdown-based knowledge base.
[COMMAND_EXECUTION]: The skill relies on the execution of generated Python scripts and filesystem commands to manage the wiki structure, rebuild indices, and process data entries.

wiki