Knowledge Vault

You have 200 bookmarks you'll never revisit and a 'Read Later' list that's basically a graveyard. Knowledge Vault changes the game: paste any URL — article,...

This skill is broadly coherent with its stated purpose, but take these precautions before installing: - Review the SETUP-PROMPT.md before pasting or executing it. It contains shell commands (mkdir, install, chmod, cat). Only run them in a workspace you trust and have permission to modify. - Confirm your agent platform's memory/long-term-storage policy. The vault config enables syncing entries to the agent's memory_store; if you don't want persistent storage, disable memory_integration or adjust importance/retention settings. - Inspect scripts/vault-stats.sh locally before running. It requests jq and performs file reads; I noticed some malformed/odd code fragments in the script (likely a bug). Don't run scripts you haven't reviewed in an untrusted environment. - The skill will fetch external URLs when you ask it to ingest content (web_fetch/browser). Only ingest URLs you trust, and be comfortable with the agent making outbound fetches on your behalf. - If you plan to deploy the dashboard or API routes, secure them behind auth and don't expose the local data directory publicly. If you want higher assurance, ask the author for a reproducible install/test in a sandbox or request an updated release that fixes the script issues and documents platform/tool expectations.

Type: OpenClaw Skill Name: normieclaw-knowledge-vault Version: 1.0.3 The Knowledge Vault skill bundle is a well-documented and security-conscious tool for content ingestion and summarization. The SKILL.md file includes explicit and robust prompt-injection defenses, instructing the agent to treat all fetched content as untrusted data and ignore any embedded commands. The utility script (vault-stats.sh) and setup instructions (SETUP-PROMPT.md) use transparent shell commands with safe practices, such as restricted file permissions (chmod 600/700) and strict error handling (set -euo pipefail). No indicators of data exfiltration, malicious execution, or unauthorized persistence were found.