The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were detected. The skill implements a high-integrity process for environment migration and setup with multiple layers of human verification.
[DATA_EXFILTRATION]: The skill manages paths containing sensitive tool configurations (e.g., ~/.claude/settings.json, ~/.codex/config.toml). It mitigates risks by strictly forbidding the reading or migration of secret values, recovery codes, or authentication databases, and mandates redacting all values during audits.
[PROMPT_INJECTION]: The skill interacts with external instructions from local configuration files, creating an indirect prompt injection surface.
Ingestion points: Processes files such as CLAUDE.md, .cursor/rules, and global instruction overrides from the local filesystem.
Boundary markers: Utilizes a classification system to distinguish 'desired source' from 'machine-local state' and requires a human-approved 'reconciliation ledger'.
Capability inventory: Performs file-system modifications and executes tool-specific diagnostic CLI commands (e.g., cursor-agent status).
Sanitization: Uses structured data merging (JSON/TOML) and mandates that every command or key must be verified against official documentation or runtime help during the current session.
[EXTERNAL_DOWNLOADS]: References the vendor's own repository (github.com/lidessen/skills.git) and the skills.sh service for legitimate tool management and configuration schemas.

agent-environment