The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses and modifies files in ~/.claude/skills/ and ~/.cursor/skills-cursor/. These directories are used by AI agents to store their global configuration and skill definitions. Although this access is necessary for the skill's synchronization functionality, these paths are considered sensitive because they control the agent's behavior across different sessions and projects.\n- [INDIRECT_PROMPT_INJECTION]: The skill provides functionality to import external skills into the repository via the --from-claude and --from-cursor flags, establishing a surface for indirect prompt injection.\n
Ingestion points: The pull_from function in scripts/sync-skills.sh reads and copies data from external global skill directories into the local repo.\n
Boundary markers: No explicit delimiters or warnings are used to isolate or mark the ingested content.\n
Capability inventory: The skill uses the Bash tool to execute a script that performs sensitive file system operations including rsync, cp, and rm.\n
Sanitization: The script verifies the existence of SKILL.md but does not perform any validation or sanitization of the contents within the files being imported.

sync-skills