The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests historical session data and user-provided artifacts to drive the evolutionary search for better prompts and code.
Ingestion points: Uses SessionExtractor to read real usage traces from agent databases (e.g., ~/.hermes/sessions.db).
Boundary markers: No specific boundary markers or instructions to ignore embedded commands within the ingested session history are documented.
Capability inventory: The skill has the ability to execute code variants, run test suites, and create Pull Requests on GitHub repositories.
Sanitization: There is no mention of sanitizing or escaping session data before it is processed by the LLM optimizer.
[DYNAMIC_EXECUTION]: The optimization pipeline involves the automated generation and execution of code mutations to evaluate performance.
Evidence: The GEPA optimization loop generates candidate variants through mutation and executes them against evaluation datasets to analyze traces.
Risk: Running code that has been dynamically mutated by an LLM could lead to unintended system side effects if executed outside of a secure sandbox.
[EXTERNAL_DOWNLOADS]: The skill instructions direct the user to download the framework from an external repository.
Evidence: Fetches the framework from the NousResearch repository on GitHub (https://github.com/NousResearch/hermes-agent-self-evolution.git).
[COMMAND_EXECUTION]: The skill relies on several shell-based commands for setup and execution.
Evidence: Instructions include the use of pip install, git clone, and python -m for running optimization and data extraction tasks.

hermes-agent-self-evolution