The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests design specifications from sdd.md files (Step 2 in planning.md). While this presents an attack surface for indirect injection, the skill implements significant mitigations: it uses explicit boundary markers (---BEGIN SDD CONTEXT---) when data is passed to sub-agents (Step 1b in create-inline-common.md), utilizes a mandatory strict validation gate (uip maestro case validate --strict) to ensure structural and semantic integrity (Step 12 in implementation.md), and employs user-in-the-loop confirmation (AskUserQuestion) for ambiguous or unresolved mappings.- [EXTERNAL_DOWNLOADS]: The skill includes automated logic to resolve the UiPath CLI binary and update it via npm install -g @uipath/cli@latest if the version is outdated (Step 0 in planning.md). These downloads originate from the official npm registry and target verified vendor packages from UiPath.- [COMMAND_EXECUTION]: The skill frequently executes shell commands via the Bash tool to interact with the UiPath CLI (uip). These commands are used for project scaffolding, registry discovery, validation, and publishing (documented in case-commands.md). Execution is restricted to official vendor tools and includes safety checks, such as requiring user consent for debug sessions that perform live API calls or emails (Rule 12 in SKILL.md).

uipath-maestro-case