Bluesky

**Quick judgement**: This is a strong, full-featured Bluesky CLI that is highly suitable for agent use, especially for reading, summarizing, and structured engagement on Bluesky. It exposes nearly all common actions (timeline, posts, threads, likes, reposts, follow/block/mute, search, notifications) with sensible, scriptable commands and JSON output for machine consumption. **Key strengths** - Broad coverage of Bluesky/AT Protocol features, including threads and media. - Consistent command naming and flags; `--json` makes parsing straightforward for agents. - Clear login flow using app passwords with a security-conscious design (password only used to obtain a session token, not stored). - Error modes and recovery steps are explicitly documented, which helps an agent self-correct. **Risks & considerations** - The skill operates directly on a real Bluesky account: it can post publicly, like/repost, follow/unfollow, and block/mute users. Misuse, prompt injection, or poorly constrained instructions could lead to spammy, reputationally harmful, or harassing behavior from the user’s account. - Reading timelines and notifications may expose private or sensitive content that the user did not intend to be processed extensively by an AI. - Session handling and logins must be done carefully; the user must understand they are granting the agent operational control of their account. **Recommended scenarios** - Power users who intentionally want an AI assistant to help manage or co-pilot their Bluesky presence: summarizing timelines, drafting posts/threads, and performing light engagement under explicit direction. - Workflows that need structured Bluesky data (search results, threads, notifications) in JSON for analysis, monitoring, or reporting. - Situations where the user is comfortable granting controlled posting and engagement rights to an agent, with appropriate guardrails (e.g., human-in-the-loop review for posts, strict policies around not following/blocking without confirmation).