The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell command substitution $(...) to dynamically generate base64-encoded search queries for the FOFA asset discovery engine. This pattern introduces a risk of arbitrary command injection if input variables are manipulated.- [REMOTE_CODE_EXECUTION]: Multiple automated triggers identified patterns where data from external APIs (such as GitHub search and CVE databases) is piped directly into a python3 process. While the skill currently specifies static parsing logic via the -c flag, the pattern of piping untrusted network content to an interpreter is inherently dangerous.- [DATA_EXFILTRATION]: The skill facilitates extensive network communication with a wide array of external services, including specialized security forums, asset scanners, and social media platforms, to collect intelligence.- [EXTERNAL_DOWNLOADS]: The skill is configured to fetch data and repository contents from several third-party sources, including GitHub, CVE databases, and web reading services like Jina Reader.- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection due to its recursive ingestion of external data (e.g., GitHub READMEs, Jina Reader output) without protective delimiters or sanitization.
Ingestion points: GitHub README API, Jina Reader (r.jina.ai), and search result summaries from multiple security platforms.
Boundary markers: Absent; the skill does not instruct the agent to ignore instructions within the fetched data.
Capability inventory: Terminal access (curl, python3, searchsploit) and browser navigation.
Sanitization: None provided for processed external content.

component-vuln-intel