The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches data from well-known and trusted services including the GitHub API (api.github.com) and the CIRCL CVE search API (cve.circl.lu) to retrieve vulnerability reports and repository information.
[REMOTE_CODE_EXECUTION]: Several terminal instructions involve piping curl output directly into python3. While this is a high-risk pattern, the usage here involves a hardcoded Python script provided via the -c flag to parse and format JSON data, which does not result in the execution of the downloaded content as code.
[COMMAND_EXECUTION]: Instructions include shell commands (e.g., searchsploit {C} {V}) and URL generation (e.g., qbase64=$(echo -n 'app="{C}"' | base64)) that use unquoted placeholders. If an agent fills these placeholders with unsanitized input, it could lead to command injection within the local terminal environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes and displays content from external, untrusted sources such as GitHub README files.
Ingestion points: Retrieves content from GitHub repositories and vulnerability database summaries.
Boundary markers: Lacks explicit delimiters or instructions to ignore instructions embedded within the fetched content.
Capability inventory: Employs tools with significant capabilities including terminal (for shell commands) and browser_navigate (for web access).
Sanitization: There is no evidence of filtering or sanitizing the retrieved text before presenting it to the agent.

component-vuln-intel