The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the shinkoku command-line utility for all ledger management and data import tasks. It invokes various subcommands such as ledger init, import csv, and ledger journal-add to interact with a local SQLite database.
[INDIRECT_PROMPT_INJECTION]: The skill serves as a processing point for untrusted external data, including transaction records from CSV files and text extracted via OCR from user-provided receipt and invoice images.
Ingestion points: Data is ingested through the shinkoku import commands and via delegation to the /reading-receipt and /reading-invoice skills.
Boundary markers: The skill contains explicit instructions that "Automatic registration is not done" and that every entry must be confirmed by the user before registration.
Capability inventory: The skill possesses the ability to execute database commands and write to files within the .shinkoku/progress/ directory.
Sanitization: While the internal logic of the CLI tool is not visible, the primary security control is the mandatory human-in-the-loop verification step for every transaction, preventing malicious data from being registered without oversight.

journal