The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute shell commands using npx to interact with the local filesystem for linting and code repair. Commands identified include npx --no-install eslint --ext .vue . and npx eslint --ext .vue . --fix. The use of --no-install for read-only checks follows the principle of least privilege by preventing the tool from downloading and executing new, unverified packages from the internet at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted local project data, which introduces a potential surface for indirect prompt injection if an attacker can influence the contents of the scanned files.
Ingestion points: The skill reads Vue source files (.vue) and multiple configuration files, including eslint.config.js, .eslintrc.json, and .pre-commit-config.yaml as part of its 'Step 2: 读取项目契约' (Read Project Contract) phase.
Boundary markers: There are no explicit delimiters or specific instructions for the agent to ignore embedded instructions within the processed files or the resulting tool output.
Capability inventory: The skill possesses the capability to execute shell commands and modify local files through the --fix operation.
Sanitization: No sanitization or content validation is performed on the ingested file contents prior to them being processed by the linting tool and subsequently reflected in the agent's context.

codeguard-vue