The Agent Skills Directory

[SAFE]: The skill is designed to work offline using a local SQLite database that is distributed with the code and opened in read-only mode. It does not perform any network operations, access sensitive file paths, or attempt to persist across sessions.
[DYNAMIC_EXECUTION]: The skill provides an 'escape hatch' through its sql command in scripts/ooxml.mjs, which allows the execution of arbitrary SQL SELECT statements against the internal schema database. This functionality is restricted to read-only queries by both the database connection settings (readOnly: true in scripts/db.mjs) and internal command validation, posing no risk to the host system.
[INDIRECT_PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection through the processing of external diagnostics.
Ingestion points: Untrusted diagnostic data enters via the explain command in scripts/ooxml.mjs.
Boundary markers: Absent; the skill expects raw JSON input.
Capability inventory: The skill has no subprocess calls, shell exec calls, file-write operations, or network operations in its scripts. All operations are limited to read-only SQL queries via the built-in node:sqlite module.
Sanitization: Present; input is parsed as JSON, and only specific fields are extracted and used for schema lookups, mitigating potential exploitation of the host.

ooxml-lookup