Blog Writer

Key things to consider before installing or enabling: - Privacy: Mode B reads sessions_history (all sessions of the day). If enabled, private chat content could be incorporated into drafted posts. Only enable if you explicitly want agent conversations summarized as blog drafts and you trust the agent environment. - Hardcoded path: scripts/new_post.py writes to /Users/nora/workspace/astralwaveorg/source/_posts. That path is specific to a user and likely incorrect for your environment — change POSTS_DIR to your repo path or make it configurable before use. - Implementation mismatch: scripts/extract_topics.py does not actually parse stdin in its main() — it prints a placeholder JSON. The SKILL.md claims the agent will exec this script and parse topics from stdout; right now that flow is broken or incomplete. Fix or validate the script logic. - Notifications: SKILL.md says to send Telegram notifications but no token/config is declared. Determine which tool will send Telegram messages and where its credentials come from; prefer explicit config variables rather than implicit global tokens. - File review & git behavior: The skill auto-creates files and runs git add + commit (no push). Review created files before deploying/pushing. Ensure the agent runs with a least-privilege account and that commits are created under an expected identity. - Test in a safe environment: Run the scripts locally in a test repo first to confirm behavior (especially extract_topics.py and new_post.py), and verify the sessions_history→extract pipeline behaves as you expect. If you want, I can: (1) point out exact lines to change to make POSTS_DIR configurable, (2) produce a corrected extract_topics.main() that actually reads stdin and outputs topics, or (3) draft a minimal wrapper that prompts for confirmation before writing/committing.