The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted LWC source code (JavaScript, TypeScript, and HTML) to provide security and compatibility reviews. This processing involves executing external tools and interpreting their output, creating an attack surface where malicious instructions in the code could attempt to influence the agent.
Ingestion points: The skill reads file content from a user-specified component bundle path as described in the Workflow (SKILL.md).
Boundary markers: While the skill provides a structured report template, it does not explicitly instruct the agent to use unique delimiters or "ignore instructions" warnings when processing the untrusted code content, which could allow malicious instructions embedded in the code to influence the agent's behavior.
Capability inventory: The skill possesses the capability to execute shell commands via scripts/run-komaci.sh and can modify files if the user requests automated fixes (Step 7).
Sanitization: The instructions focus on mapping tool output to provided remediation text, which serves as a control, though no formal sanitization of the input code is performed before processing.
[COMMAND_EXECUTION]: The skill executes a bundled shell script scripts/run-komaci.sh to run the Komaci analyzer. The script is designed with safety controls, including argument quoting and file existence checks, to prevent command injection from user-provided paths.
[EXTERNAL_DOWNLOADS]: The skill downloads and installs Node.js dependencies (eslint and @salesforce/eslint-plugin-lwc-graph-analyzer) from the public npm registry. These are official plugins from a well-known vendor (Salesforce) and are pinned to specific versions to ensure consistency and safety.

mobile-platform-offline-validate