Podwise

Podcast knowledge workflows powered by Podwise CLI: search podcasts and episodes by keyword, monitor followed shows for new releases, find popular episodes,...

What this means and what to check before installing/use: - This is an instruction-only skill that drives the Podwise CLI. It does not bundle or execute code itself; instead it runs podwise commands and reads/writes files in the current working directory. That behavior aligns with the described podcast workflows. - Installation: the included installation doc suggests several install options. The automatic curl | sh installer (raw.githubusercontent.com/hardhackerlabs/...) and an untrusted Homebrew tap are the riskiest suggestions — do NOT run curl | sh without reviewing the script first. Prefer installing from a GitHub release archive you inspect, building from source, or using a package manager you trust. - Local file access: the skill will read a local taste.md for personalization and will write note/export files to your current working directory (e.g., debate summaries, episode notes, Anki/CSV exports). If taste.md may contain sensitive data, review or remove sensitive fields before using the skill. - Config and API keys: Podwise CLI stores config in ~/.config/podwise/config.toml and requires an API key or browser auth. Confirm any API keys are from podwise.ai and that you trust that service. Exports to third-party services (Notion, Readwise) will use whatever credentials the Podwise CLI or its auth flow provides — verify and control those integrations in the CLI before exporting. - Quota/processing: running podwise process consumes Podwise credits. The workflows explicitly require confirming before processing; the agent should ask you before using quota. Always confirm before allowing processing of remote content or local files. - Transparency: some workflow steps say to 'read' taste.md silently and to 'silently record' internal notes used for a later saved summary. This is not exfiltration, but it means the agent will access and temporarily store some internal state. If you want full transparency, ask the agent to show any notes it recorded before saving them to disk or exporting. If you decide to proceed: (1) review the install script before running it, (2) inspect ~/.config/podwise/config.toml and any connected integrations, (3) ensure taste.md contains only the non-sensitive preferences you want the agent to use, and (4) be prepared to confirm any processing that consumes credits or exports data externally.

Type: OpenClaw Skill Name: podwise Version: 0.2.0 The skill bundle facilitates interaction with the Podwise podcast service but contains high-risk patterns and potential vulnerabilities. Multiple workflows, such as 'episode-notes.md' and 'topic-research.md', instruct the AI agent to construct and execute shell commands using potentially unsanitized user-provided strings (e.g., episode titles, search keywords, and file paths), creating a significant risk for shell injection. Furthermore, 'references/installation.md' encourages a high-risk 'curl | sh' installation method (pointing to raw.githubusercontent.com/hardhackerlabs/podwise-cli/main/install.sh). While these capabilities are plausibly intended for the stated purpose, the lack of input sanitization instructions and the use of risky installation scripts warrant a suspicious classification.