The Agent Skills Directory

[COMMAND_EXECUTION]: The skill implements a command-line interface (bin/cli.mjs) to perform PII detection and redaction locally. This is a core feature of the skill and is executed using the Node.js runtime without external network calls.
[INDIRECT_PROMPT_INJECTION]: The skill processes external text files provided by the user. While ingesting untrusted data is a potential vector for indirect prompt injection, the skill acts as a security tool to sanitize this data and includes a residual PII detection gate (src/audit-log.mjs) to prevent the accidental inclusion of sensitive information in the output.
Ingestion points: User-specified files or stdin processed via bin/cli.mjs.
Boundary markers: None explicitly implemented for the LLM interaction; however, the processing logic is deterministic.
Capability inventory: Local file system read/write access and execution of the provided Node.js scripts via the Bash tool.
Sanitization: Implements mandatory verification via the ResidualPIIError logic to ensure no original PII values are present in the final output.
[DATA_EXFILTRATION]: Analysis of the source code confirms the tool is designed for offline use. It uses standard Node.js modules for local file processing and cryptography, and does not perform any network operations to transmit data externally. The AuditLog purposefully excludes PII values to ensure compliance with privacy principles.

let-it-be