The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from .claude/settings.local.json, which is an untrusted project-local file. This creates a surface for indirect prompt injection where malicious instructions embedded in settings values could attempt to influence the agent's behavior during the diffing and confirmation workflow.
Ingestion points: Reads .claude/settings.local.json from the current project root (SKILL.md).
Boundary markers: None present to delimit the untrusted JSON content from instructions.
Capability inventory: The skill has the capability to read from and write to the filesystem, specifically modifying the global ~/.claude/settings.json file.
Sanitization: No explicit sanitization of setting values for injection patterns is performed, though the skill suggests generalization of paths.
[COMMAND_EXECUTION]: The skill performs filesystem operations to read and update sensitive configuration files, including ~/.claude/settings.json and .claude/settings.local.json. This capability is used to manage the agent's global configuration and permissions as requested by the user.
[EXTERNAL_DOWNLOADS]: The README.md provides installation instructions that fetch the skill from the shihyuho/skills repository using the npx skills add command.

promote-claude-settings