The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external web application files and DOM content during the inventory and verification loops, creating a potential surface for indirect prompt injection. The skill mitigates this by providing specific security instructions in references/security.md that mandate output character limits, the use of boundary markers, and the redaction of sensitive information before it enters the model context. \n
[EXTERNAL_DOWNLOADS]: The Workbench utility (scripts/workbench.mjs) is capable of downloading the Playwright library and browser binaries from official registries. These downloads are necessary for the skill's primary verification functionality, are restricted to specific pinned versions, and require explicit user approval via command-line flags or interactive confirmation prompts. \n
[COMMAND_EXECUTION]: The skill executes standard system commands such as npm, flock, and lockf using the Node.js spawn method. These executions are scoped to project-internal operations like dependency installation and implementing an advisory file-locking protocol to ensure the integrity of the manifest during concurrent operations.

webmcpify