The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data by auditing the project's codebase and configuration files (such as composer.json and .env) to determine upgrade steps. This creates a surface for indirect prompt injection where malicious content within the audited files could attempt to influence the agent's instructions.
Ingestion points: composer.json, .env, and PHP source files via the fuse-ai-pilot:explore-codebase agent.
Boundary markers: None explicitly defined to separate audited code content from agent instructions.
Capability inventory: composer commands, bash script execution, and local file writes.
Sanitization: No explicit sanitization or validation of the ingested code content is mentioned before processing.
[COMMAND_EXECUTION]: The skill provides a bash script template in references/templates/upgrade-script.sh.md and instructs the user/agent to execute it. This script performs several system-level actions:
Appending configuration values to the .env file.
Running composer update to modify project dependencies.
Executing php artisan commands to clear application caches.
[DYNAMIC_EXECUTION]: The skill involves the generation and execution of a local shell script (upgrade-laravel-13.sh) based on a provided template. This is used to automate the transition between Laravel versions by chaining multiple shell and PHP commands.
[EXTERNAL_DOWNLOADS]: The instructions direct the agent to fetch and update numerous external dependencies via the Composer package manager. These include core framework components and testing utilities from well-known sources in the Laravel and PHP ecosystems, such as laravel/framework, phpunit/phpunit, and pestphp/pest. The skill also references external documentation and source code repositories on GitHub for manual review.

laravel-upgrade-v13