The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation provides instructions for downloading the Podwise CLI tool from vendor-hosted locations, including a Homebrew tap, an installation script on GitHub, and binary releases from the hardhackerlabs repository.- [REMOTE_CODE_EXECUTION]: The installation workflow includes the standard pattern of piping a remote shell script from GitHub directly to the system shell for execution. This is the documented installation method for the vendor tool and is consistent with the skill's setup requirements.- [COMMAND_EXECUTION]: The skill executes multiple subcommands using the podwise utility to fetch data, process media, and manage configuration. It also involves shell commands for file management and permissions during the installation phase.- [CREDENTIALS_UNSAFE]: The skill references the storage of authentication tokens in ~/.config/podwise/config.toml and provides command examples for setting these API keys manually. It interacts with these credentials to perform authenticated actions.- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and summarize external audio transcripts and summaries from sources like YouTube and RSS feeds. This presents a potential attack surface where instructions embedded in transcript text could be interpreted by the LLM. Evidence: 1. Ingestion points: Transcripts and summaries fetched via the CLI. 2. Boundary markers: Not explicitly implemented in prompt logic. 3. Capability inventory: CLI execution and file writing. 4. Sanitization: Not explicitly described in instructions.

podwise