The Agent Skills Directory

[DATA_EXPOSURE_AND_EXFILTRATION]: The skill accesses the global configuration file ~/.claude/settings.json and the project-specific configuration file .claude/settings.local.json. These files define the agent's security policies and allowed tools.- [PRIVILEGE_ESCALATION]: The skill modifies .claude/settings.local.json to add new permissions. This allows the agent to programmatically expand its own set of allowed capabilities within a project context.- [INDIRECT_PROMPT_INJECTION]: The skill processes data from configuration files to determine its execution environment.
Ingestion points: Reads ~/.claude/settings.json and .claude/settings.local.json (SKILL.md).
Boundary markers: None; the skill assumes the content of these files is trusted.
Capability inventory: Includes Write and Edit permissions, and shell command execution via Bash tools.
Sanitization: None; the permissions found in the configuration files are added directly to the local settings.

sync-permissions