The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill contains instructions to fetch content from an external URL (https://codex-reset.com/api/timeline) and pipe it directly to a Python interpreter. Although the provided script is intended for JSON parsing, this execution pattern is a high-risk vector that allows a remote server to potentially execute arbitrary code on the agent's system.\n- [EXTERNAL_DOWNLOADS]: The skill initiates network requests to codex-reset.com, which is identified as a malicious or blacklisted domain by security scanners. It also retrieves data from other third-party services like r.jina.ai and x.com.\n- [COMMAND_EXECUTION]: The skill provides various shell commands for the agent to run, including complex date manipulations and network fetching tools, which increase the attack surface of the skill execution environment.\n- [INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted data from community-run websites and social media platforms to inform its conclusions. Malicious instructions hidden in these external sources could influence agent behavior, as the implementation lacks boundary markers, sanitization, or input validation.\n
Ingestion points: Fetches from codex-reset.com, r.jina.ai (scraping x.com), and general web search.\n
Boundary markers: None identified in the prompt logic to separate system instructions from external data.\n
Capability inventory: Includes curl for network access and python3/bash for local script execution.\n
Sanitization: No filtering or escaping mechanisms are applied to the external content before it is processed by the agent.

tibo-reset-codex