The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands including git (diff, status, ls-files) and the evo CLI to manage repository state and execute benchmarks. These operations are performed locally and are essential to the skill's primary function of code optimization.
[EXTERNAL_DOWNLOADS]: The skill references the installation of evo-hq-cli and evo-hq-agent (via uv, pip, or npm). These are official resources belonging to the vendor 'evo-hq'. The skill explicitly instructs the agent to ask for user confirmation before performing any installation and provides manual installation steps for the user.
[DATA_EXPOSURE]: The skill provides instructions for loading local .env files using evo env load. This is a standard administrative feature for providing benchmarks with necessary runtime credentials and is managed locally by the vendor's CLI.
[DYNAMIC_EXECUTION]: The skill involves the creation and execution of benchmark harness scripts within isolated experiment worktrees. This runtime code generation and execution is the core purpose of the optimization tool.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill reads local repository files such as READMEs, entry points, and configuration files to understand the codebase (SKILL.md Section 1).
Boundary markers: Not explicitly specified during the repository exploration phase.
Capability inventory: The skill can execute shell commands and modify the filesystem within dedicated experiment worktrees.
Sanitization: No explicit sanitization of repository content is mentioned, but the data is used to inform the agent's choreography rather than as direct command input.

discover