The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes various local shell commands and scripts to manage the documentation lifecycle.
Evidence:
Executes shasum -a 256 via Bash to perform incremental translation checks based on file hashes.
Runs node skills/doc-smith-build/scripts/build.mjs to rebuild the navigation structure (nav.js) after translation.
Uses standard utilities like ls, wc, cat, mkdir, and rm for workspace maintenance and status polling.
These commands target vendor-owned scripts and internal project directories (.aigne/doc-smith/).
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to its core function of processing and translating untrusted documentation content.
Ingestion points: The agent reads HTML files from dist/{sourceLanguage}/docs/ to extract translatable text.
Boundary markers: The workflow utilizes specific HTML structural markers (<main data-ds="content">, <nav data-ds="toc">) to isolate translatable areas from the rest of the document.
Capability inventory: The skill possesses file read/write capabilities, Bash execution for integrity checks, and the ability to trigger sibling skills and local Node.js scripts.
Sanitization: The instructions include strict constraints to preserve existing HTML tags, skip code blocks (<pre><code>), and explicitly forbid the introduction of new <script> tags or event attributes during the translation phase.

doc-smith-localize