The Agent Skills Directory

[PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill is designed to ingest and process untrusted data from the internet to ground its analysis, which can be exploited to influence the agent's behavior.
Ingestion points: External data enters the agent's context via the WebSearch tool during the "Elon Musk Research" phase (Step 2) in SKILL.md.
Boundary markers: The skill lacks explicit boundary markers (such as XML tags or dedicated markdown wrappers) or instructions to ignore embedded commands within the retrieved search results.
Capability inventory: The agent utilizes WebSearch and local file reading across various research documents in the references/ directory.
Sanitization: There are no instructions for sanitizing, escaping, or validating the content retrieved from external URLs before it is processed by the LLM.
[SAFE]: Persona Management and Constraints. The skill defines a clear role-play framework in SKILL.md that includes a mandatory one-time disclaimer and explicit exit triggers ("退出", "切回正常") to allow the user to terminate the persona easily.
[SAFE]: External Resources and Installation. The skill documentation provides installation instructions using standard package managers (npx) and version control (git). The external links and repositories trace back to the author (alchaincyf) or well-known services (vercel-labs/skills), representing standard developer functionality for this ecosystem.

elon-musk-perspective