The Agent Skills Directory

[DATA_EXFILTRATION]: The skill is designed to read and process highly sensitive files and directories to perform its function.
Evidence:
Specifically targets and reads ~/.claude/sensitive-content-context.md to load user-defined sensitive patterns.
Scans for credentials in .env files, .ssh keys, AWS credentials (~/.aws/credentials), and other sensitive local paths mentioned in the detection categories.
While this behavior is consistent with the primary purpose of a security scanner, it brings sensitive system information into the agent's context.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted content from user-provided file paths, creating a vulnerability surface where malicious instructions could be embedded in the data being scanned.
Ingestion points: Arbitrary files located in the path provided to the /sensitive-content-scanner command as described in SKILL.md.
Boundary markers: No specific delimiters or "ignore embedded instructions" warnings are defined for the content being scanned or reported.
Capability inventory: The skill utilizes file system read access to ingest data and generates a structured markdown report containing snippets of the scanned files.
Sanitization: No explicit sanitization or escaping of the scanned content is mentioned before it is interpolated into the findings report.

sensitive-content-scanner