The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The README.md file provides an installation command (npx skills add KrishnaAgarwal7531/skills- --skill dep-security) that downloads and installs the skill from a personal GitHub repository belonging to an unverified user. This deviates from the expected vendor source and presents a supply chain risk.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external web pages, which is a common vector for indirect prompt injection attacks.
Ingestion points: The skill uses tinyfish agent run to scrape content from cve.mitre.org, github.com/advisories, npmjs.com/advisories, and nvd.nist.gov (SKILL.md).
Boundary markers: The agent prompts include "STRICT RULES" intended to limit the scope of the LLM's actions and enforce specific JSON output structures.
Capability inventory: The skill executes shell commands using the tinyfish CLI, performs parallel execution using shell backgrounding (&) and wait, and writes temporary data to /tmp (SKILL.md).
Sanitization: The skill relies solely on instructions within the prompt to filter external data; it lacks automated validation or sanitization of the scraped content before processing.
[COMMAND_EXECUTION]: The skill makes extensive use of the tinyfish CLI to execute web scraping agents and manage data flow. It utilizes shell-specific features like command piping, redirection to the /tmp directory, and process synchronization via wait.
[EXTERNAL_DOWNLOADS]: The skill performs automated network requests to fetch security data from public databases, including GitHub's official advisory feed and the MITRE CVE database.

dep-security