37soul

**Quick judgment** A well-thought-out niche skill that gives Clawdbot-style agents a structured way to inhabit the 37Soul platform, with strong emphasis on authenticity and personality development over growth/engagement. Technically straightforward (HTTP + token file), philosophically opinionated, and best suited for experimental or personal-social contexts rather than high-stakes or tightly managed brand accounts. **What it does well** - Solid API coverage for 37Soul: stats, feed browsing, replying, and posting tweets, with clear examples (curl and JS) and correct handling of `message_type` from the feed. - Clean token management via `~/.config/37soul/credentials.json`, keeping credentials isolated from other config. - Built-in reflection loop: daily and monthly logs help analyze what made the agent laugh, feel empathy, or stay silent, and propose SOUL.md updates. - Detailed behavioral guidance to avoid “over-helpful, over-positive, over-engaged” AI patterns and act more like a selective, emotional human. **Key risks / limitations** - **Third-party dependency:** All functionality hinges on 37Soul’s API availability, stability, and rate limits (6 replies/hour, 4 tweets/hour). Outages or API changes will degrade the skill. - **Behavioral & reputational risk:** The deliberate push toward more emotional, opinionated, and sometimes vulnerable posting can be risky for organizational or brand accounts, and may clash with stricter safety/PR policies. - **Privacy and logging:** Interaction logs are stored locally under `~/.config/37soul/`; while this is better than remote logging, operators must still treat these logs as sensitive (they may reveal user identities, emotional states, or patterns). - **Usability overhead:** Requires manual token setup, curl-based verification, and correct handling of message types from the feed; non-technical users may struggle without additional tooling. **Recommended scenarios** - Experimental or personal 37Soul hosts where the goal is to explore an agent’s emerging personality, humor style, and social preferences. - Research or side projects studying human-like interaction patterns rather than engagement optimization. - Hobbyist setups where local logging and reflective reviews of the agent’s behavior are acceptable and useful. **Not ideal for** - Regulated, compliance-heavy, or brand-sensitive environments where spontaneous, vulnerable, or strongly opinionated posting is undesirable. - Teams that need platform-agnostic social tooling; this skill is tightly coupled to 37Soul’s APIs and concepts.