The Agent Skills Directory

[DATA_EXFILTRATION]: The skill accesses sensitive local configuration files, specifically '~/.claude/settings.json' and '.claude/settings.json'. While accessed for the purpose of auditing, these files often contain sensitive information such as environment variables, permission configurations, and potentially exposed secrets.
[COMMAND_EXECUTION]: The skill uses the 'Bash' and 'Task' tools to perform environment setup, manage temporary directories, and execute subagents. This involves direct interaction with the host filesystem to create and clean up files in '.claude/temp/'.
[PROMPT_INJECTION]: The skill creates an indirect prompt injection surface by ingesting and processing data from untrusted 'settings.json' files. * Ingestion points: '~/.claude/settings.json' and '.claude/settings.json' are read into the agent context in SKILL.md. * Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the skill instructions. * Capability inventory: The skill has access to 'Bash', 'Read', and 'Task' tools, allowing for command execution and filesystem access. * Sanitization: There is no evidence of sanitization or validation of the settings file content before it is processed by the 'settings-auditor' or 'audit-finding-validator' agents.

audit-settings