The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function involves reading and analyzing project source code, routing configurations, and component libraries to integrate new tools. This creates an attack surface where instructions hidden within the analyzed project files could potentially influence the agent's code-generation or modification behavior.
Ingestion points: Reads local project files, configuration files, and existing codebases as specified in the workflow.
Capability inventory: The skill is authorized to perform file writes (components, sitemaps, translations), modify project navigation, and execute build/test commands.
Boundary markers: The skill instructions emphasize keeping analysis and preview requests read-only, separating them from actual modifications.
Sanitization: The skill relies on user review of delivered code and validation evidence rather than automated input sanitization.
[COMMAND_EXECUTION]: The skill utilizes Python scripts for management tasks and expects to execute shell commands for project builds and type checking.
Evidence: The skill includes scripts/profile_store.py for state management and scripts/validate_skill.py for integrity checks. The workflow in SKILL.md describes executing build and test suites.
Safety controls: The profile_store.py script implements a blacklist (e.g., token, secret, api_key) to prevent the persistence of sensitive credentials within the shared user profile.

lov-add-tool