remix2111/hermes-agent-advanced-skills — reverse-engineered prompt
Reverse engineered prompt
Build me a Hermes Agent advanced skills package that makes delegated sub agents smarter and safer.
I want one skill that automatically checks the target folder before a delegate task runs, reads an AGENTS.md or .hermes-rules file if it exists, and adds those rules into the sub agent context so it follows the right local conventions for that directory.
I also want a failure auto replan skill. When a delegate task fails, it should look at the failure reason, decide if it’s worth retrying, change the approach once, and then either succeed or return a clear failure report showing the original error, the retry attempt, and the final result. Don’t let it loop forever.
Include a simple personality called smart replan that enables the behavior, a package manifest, install instructions, and examples for project level config with .hermes-agent.yaml. Keep it usable as a drop in Hermes skill package without modifying Hermes itself.
Want more depth? Deep Reverse