The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/install.sh script and installation instructions perform automated downloads of external software, including the slides-grab package from npm and Chromium via Playwright.
[REMOTE_CODE_EXECUTION]: The skill executes code downloaded from external sources during the installation process and through runtime operations like slides-grab edit, which launches a local server.
[COMMAND_EXECUTION]: The skill uses the Bash tool to execute various CLI commands for slide generation, validation, and export, including slides-grab, yt-dlp, and npx commands.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. Attackers could embed malicious instructions in the source topics, notes, or external assets (images/videos) that the agent processes and renders into slide content.
Ingestion points: User-provided topic notes, external assets fetched via WebFetch, and images generated via external providers.
Boundary markers: None detected in the instructions to prevent the agent from following instructions embedded in the slide HTML.
Capability inventory: The skill has Bash execution, Read/Write file access, and WebFetch network capabilities.
Sanitization: No evidence of sanitization for the HTML/CSS content generated from untrusted inputs.
[CREDENTIALS_UNSAFE]: The skill interacts with sensitive authentication files, specifically ~/.codex/auth.json, and requires users to provide OPENAI_API_KEY or GOOGLE_API_KEY for specific image generation providers.
[REMOTE_CODE_EXECUTION]: The default image generation provider, god-tibo-imagen, utilizes an 'unsupported private Codex backend', which represents an opaque and potentially insecure execution environment.

slides-grab