The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes source code from the project's repository during the audit and refactoring phases, creating an attack surface for instructions embedded in data.
Ingestion points: The scripts/audit_node24.sh script reads various source files (.js, .ts, .json) using grep/ripgrep, and the fix subagent is prompted to read handler code end-to-end as defined in references/fix-brief.md.
Boundary markers: While markdown code blocks are used to delimit code, there are no specific system instructions or delimiters that command the agent to ignore potentially malicious prompts embedded within the source code files themselves.
Capability inventory: The orchestrator agent has the capability to write to critical configuration files (Amplify backend configs, package.json), install packages via npm, and execute shell scripts. The subagents have permission to modify function source code.
Sanitization: The skill does not perform sanitization, filtering, or escaping of the ingested code content before it is processed by the AI models.
[COMMAND_EXECUTION]: The skill relies on shell command execution for its primary functionality.
Evidence: The workflow involves running a local shell script (scripts/audit_node24.sh) to identify breaking changes across the codebase.
Evidence: The instructions direct the agent to use standard development tools such as npm, npx, amplify push, and tsc for dependency management, deployment, and build verification.

amplify-lambda-node24-upgrade