The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several local bash scripts (scaffold-docs.sh, new-volume.sh, audit-docs.sh, rename-screenshots.sh) to manage the documentation lifecycle. These scripts perform various file system operations, including copying templates, moving files (using git mv where applicable), and modifying file contents in-place using sed. While these are standard developer utilities, they execute shell commands to modify the local workspace environment.
[INDIRECT_PROMPT_INJECTION]: The primary function of the skill is to read and interpret legacy source code (Java/GBK) to generate structured documentation for a new system architecture. This creates an attack surface where malicious or specially crafted content within the analyzed legacy codebase could potentially influence the agent's reasoning or the generated documentation outputs.
Ingestion points: The skill reads files from a legacy source code repository (path defined in .agents/paths.env).
Boundary markers: No explicit markers are used to separate ingested code data from system instructions during the documentation generation process.
Capability inventory: The skill includes scripts that can read, write, and rename files within the documentation directory, and it has the capability to execute system utilities like iconv, sed, awk, and git via shell scripts.
Sanitization: The skill performs character encoding conversion (GBK to UTF-8) using iconv, but it does not perform semantic sanitization of the ingested code content before it is processed by the model.

kthis-migration-plan