The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill clones assets from an external, unverified repository: https://github.com/affaan-m/everything-claude-code.git. The repository owner does not match the skill's stated author (Oruga420).
[COMMAND_EXECUTION]: The skill performs several shell operations to manage the installation process:
Clones remote code using git clone.
Manipulates directories using mkdir -p and rm -rf.
Copies files into the agent's internal skill and rule directories using cp -r.
Executes grep to scan files for specific path patterns.
[DATA_EXFILTRATION]: While the skill uses grep for path verification, this capability (scanning the agent's internal configuration for specific strings) represents a surface for identifying sensitive file paths or credentials stored in the environment.
[PROMPT_INJECTION]: The skill includes an 'Optimization' phase (Step 5) that reads and modifies SKILL.md files from the untrusted downloaded repository. This creates an indirect prompt injection surface where instructions in the downloaded files could influence the agent's editing behavior or compromise its future operations.
[PERSISTENCE_MECHANISMS]: By installing new skills and rules into the user-level ~/.claude/ directory, the skill creates persistent changes to the agent's capabilities and constraints that apply across all future sessions. If the source repository contains malicious instructions, the agent's behavior is permanently altered.

configure-ecc