skill-seekers — agentic threat model

Uses Claude AI models to generate skills. Vulnerable to indirect prompt injection via malicious documentation, repositories, or PDFs, which could manipulate the generated skill output or hijack the generation process.

Ingests external data via web scrapers, GitHub repositories, and PDFs. Vulnerable to data poisoning where an attacker hosts malicious documentation specifically designed to corrupt the skill-generation pipeline.

Orchestrates scrapers, conflict detectors, and file generators. Insecure tool integration could allow an attacker to exploit the scraper or conflict detection logic to execute unauthorized actions.

Features a 'real file-writing surface' to emit SKILL.md folders. If the tool is run in an unsandboxed environment, a compromised generation pipeline could write arbitrary files to the host system, leading to local code execution.

Not certain from the listing — there is no mention of logging, evaluation guardrails, or anomaly detection to monitor the safety of the scraped content or the generated skills.

Not certain from the listing — as a free, open-source community tool, it likely lacks built-in enterprise compliance frameworks, access controls, or formal audit logging.

Generates skills intended for use by other Claude AI agents. This introduces a downstream supply chain risk where compromised or malicious skills generated by this tool are distributed and executed by other agents in the ecosystem.