The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from various external web sources (OpenAI, Anthropic, Google, GitHub, etc.) to generate newsletter summaries and deep dives. This creates a surface for indirect prompt injection, where malicious instructions hidden in external web pages or repository descriptions could influence the agent's generated output or behavior.
Ingestion points: The gather-ai-news.ts script uses the WebSearch tool to query and fetch content from URLs defined in references/ai-news-sources.md.
Boundary markers: The skill documentation does not mention the use of delimiters or 'ignore instructions' warnings when interpolating external content into the synthesis prompts.
Capability inventory: The skill possesses the capability to write files to the local filesystem (newsletter HTML/text files) and perform network operations (uploading images to a CDN).
Sanitization: There is no explicit description of sanitization, filtering, or validation of the fetched external content before it is processed by the model.
[COMMAND_EXECUTION]: The workflow relies on executing several TypeScript scripts using the npx tsx runner to perform data gathering, research indexing, and newsletter generation.
Evidence: SKILL.md documents several script execution patterns: npx tsx scripts/gather-ai-news.ts, process-research.ts, and generate-newsletter.ts.
[EXTERNAL_DOWNLOADS]: The skill is configured to fetch updates and content from a wide array of AI research laboratories and software repositories.
Evidence: references/ai-news-sources.md lists official blogs and repositories from trusted organizations including OpenAI, Anthropic, Google DeepMind, Meta AI, Vercel, and others. These downloads are performed to gather intelligence for the newsletter and are consistent with the skill's stated purpose.

ai-architect-newsletter