Proof Engine

Transforms every result [PRINCIPAL_NAME] achieves into deployable proof across all business domains. Captures P&L, agent performance, funnel revenue, testimo...

Things to check before installing: - Confirm the registry-level metadata vs SKILL.md: the package claims no required env vars but both SKILL.md and README require TELEGRAM_BOT_TOKEN and TELEGRAM_CHAT_ID. Ask the publisher to clarify. - Review the included proof_manager.py yourself (or have someone trusted do so). It performs wide workspace reads (including /workspace/memory and revenue folders) and will auto-capture data. Ensure there are no secrets or credentials in those locations you don't want exposed to the skill. - If you need Telegram notifications, create a dedicated bot/chat and limited token (don’t reuse sensitive bot tokens) and test in an isolated sandbox workspace first. - Because the skill can run autonomously and scans many directories, avoid installing it in a production environment containing sensitive keys or private data until you are confident of its behavior. - If you proceed, set PROOF_DIR to an explicit, narrow path you control and verify resulting writes (vault, dashboard, stories). Monitor network calls during initial runs to confirm only the expected Telegram API endpoint is used. - The skill source appears to implement the described functionality (not obviously malicious), but the combination of broad read scope + network notifications + metadata mismatch warrants caution.

Type: OpenClaw Skill Name: proof-engine Version: 1.0.1 The proof-engine skill is a business automation utility designed to aggregate performance metrics (P&L, revenue, and testimonials) from various workspace directories and transform them into marketing content. The core logic in proof_manager.py facilitates data collection from files like tracker_state.json and AUDIT.md, manages a local financial dashboard, and generates markdown-based storytelling arcs. It uses the Telegram Bot API for notifications as explicitly declared in the SKILL.md metadata, and no evidence of data exfiltration, malicious execution, or unauthorized persistence was found.