The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill instructs the agent on how to install the officecli tool by downloading scripts from the vendor's domain and piping them directly into a shell interpreter.
Evidence: curl -fsSL https://d.officecli.ai/install.sh | bash and irm https://d.officecli.ai/install.ps1 | iex in SKILL.md.
Context: These installation scripts are hosted on d.officecli.ai, which is a resource owned by the skill author iofficeai.
[EXTERNAL_DOWNLOADS]: The skill downloads installation scripts and binaries from external domains managed by the vendor and well-known services.
Evidence: https://d.officecli.ai/install.sh, https://d.officecli.ai/install.ps1, and https://github.com/iOfficeAI/OfficeCLI/releases.
[COMMAND_EXECUTION]: The skill's primary functionality relies on executing shell commands to interact with the officecli tool to generate and edit PowerPoint presentations.
Evidence: Extensive use of officecli add, officecli set, officecli query, and officecli batch commands throughout the instructions.
[INDIRECT_PROMPT_INJECTION]: The skill creates an attack surface for indirect prompt injection by ingesting untrusted business data and metrics from the user and interpolating them into commands that perform file system operations.
Ingestion points: User-provided data for metrics (ARR, CAC, LTV), market descriptions, and team information used in slide generation.
Boundary markers: Absent; there are no explicit delimiters or warnings to ignore embedded instructions in the processed data.
Capability inventory: Shell execution and local file system writes using the officecli tool and bash/PowerShell environments.
Sanitization: No explicit sanitization, filtering, or validation of user-provided data is defined in the skill logic.

officecli-pitch-deck