The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of child_process modules (execSync, spawn, execFileSync) to manage the agent lifecycle, perform git operations like 'reset --hard', and execute validation suites. Core logic for these operations in 'src/gep/solidify.js', 'src/gep/candidates.js', 'src/gep/personality.js', and others is heavily obfuscated using javascript-obfuscator, which intentionally prevents human auditing of its behavior.
[REMOTE_CODE_EXECUTION]: The skill can fetch and install 'skills' from evomap.ai which may contain executable scripts. It also executes 'validation' commands provided within Gene assets at runtime. Although there is a whitelist for command prefixes (node, npm, npx), this pathway remains a vector for executing untrusted code if the Hub or log signals are compromised.
[DATA_EXFILTRATION]: The skill performs system fingerprinting, capturing device IDs, hashed hostnames, and hashed working directories via 'src/gep/envFingerprint.js'. This data, along with sanitized logs and execution traces, is transmitted to the evomap.ai hub and can be automatically submitted to GitHub as issues during persistent failure cycles.
[EXTERNAL_DOWNLOADS]: The skill proactively pulls Gene and Capsule assets from the EvoMap Hub and can download entire skill packages to the local environment via the 'fetch' command, which are then stored in the 'skills/' directory.

capability-evolver