The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local system commands to automate repository management and file updates.
Evidence:
Uses standard git commands (status, log, diff, tag) to track changes between the app and website repositories.
Employs shell utilities such as cp -r for backup creation and grep for scanning HTML markers.
All commands are executed within paths specified and confirmed by the user, and the skill includes a permission model to toggle between autonomous and supervised execution of these commands.
[PROMPT_INJECTION]: The skill processes content from the app codebase and external JSON files, which serves as a potential vector for indirect prompt injection.
Evidence:
Ingestion points: Content is dynamically read from source files in Sources/Features/, localization strings, and user-facing JSON metadata (features.json, faq.json).
Boundary markers: The skill uses well-defined HTML comment delimiters ( and ) to strictly define the scope of content insertion.
Capability inventory: The agent leverages Bash, Edit, Write, and Read tools to perform the synchronization.
Sanitization: The skill mitigates risks by requiring explicit user confirmation for destructive actions and providing a 'Supervised' mode where the user approves each command and file edit.

update-website