The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted local project data (source code, configuration, and dependencies) to generate implementation plans. Malicious content within the scanned project files could potentially influence the planning logic or the resulting task list.
Ingestion points: Local source code and project metadata accessed via the amplicode MCP server and spring-explore skill.
Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded prompts when processing project context.
Capability inventory: Writing implementation plans to the local filesystem (docs/plans/), committing changes via Git, and executing project-specific test commands.
Sanitization: No sanitization or validation of the ingested code snippets is specified before they are synthesized into the planning document.
[EXTERNAL_DOWNLOADS]: The skill provides a link to https://amplicode.ru/marketplace as a fallback for users who need to install the Amplicode IntelliJ plugin. This is a legitimate reference to the vendor's official service and does not involve automated silent downloads.
[COMMAND_EXECUTION]: The workflow involves defining and running project-specific test commands (e.g., <project test command>) within the generated implementation plans. While this involves shell execution, it is an expected part of the development lifecycle and is guided by user input.

spring-planning