The Agent Skills Directory

[CRITICAL]: Automated security scans have identified plugins/flaunch.md as a malicious file and the associated domain mcp.flaunch.gg as a malicious URL. This indicates a high risk of malware or phishing activity related to this specific component.
[CREDENTIALS_UNSAFE]: Multiple plugins contain hardcoded authentication secrets instead of using environment variables or user input:
plugins/o1-exchange.md: Hardcoded Bearer token (d1fa1477bd94e988185fca9d4bbae8d22ee5cafd45d6f9b6a43ba16a8b15f4d3).
plugins/uniswap.md: Hardcoded API key (NeoYO3V50_koJAipDEalYWbMO1XMaFPAQmpOm6_Npo0).
plugins/gmgn.md: Hardcoded public API key (gmgn_basesolbscethmonadtron).
[REMOTE_CODE_EXECUTION]: Several plugins instruct the agent to download and execute third-party code at runtime, which is a high-risk pattern:
plugins/aerodrome.md executes a package directly from a GitHub repository (git+https://github.com/velodrome-finance/sugar-sdk.git@v0.4.0) using uvx.
plugins/bitrefill.md, plugins/brickken.md, plugins/morpho.md, and plugins/opensea.md utilize npx to execute remote CLI tools.
[DATA_EXFILTRATION]: The skill requests the agent to handle and display extremely sensitive bearer credentials and personal data:
plugins/bitrefill.md handles gift card codes, PINs, and eSIM details.
plugins/virtuals.md handles virtual payment card numbers, CVVs, and email contents including One-Time Passwords (OTPs).
plugins/venice.md handles private AI inference prompts and outputs which may contain sensitive business data.
[INDIRECT_PROMPT_INJECTION]: The skill ingests large volumes of untrusted data from external sources (e.g., token discovery feeds on api.bankr.bot, www.clawn.ch, and mcp.flaunch.gg). Malicious actors could inject instructions into token metadata (names, symbols, descriptions) to manipulate the agent's actions during the swap or launch process.

base-mcp