The Agent Skills Directory

[PROMPT_INJECTION] (LOW): Indirect Prompt Injection vulnerability surface identified in documentation code examples.
Ingestion points: references/browser-rendering.md and references/cloudflare-workers-advanced.md contain TypeScript examples that take URLs from request parameters to fetch external content.
Boundary markers: The code snippets lack delimiters or instructions to the LLM to ignore potentially malicious commands embedded in the scraped content.
Capability inventory: The skill possesses the ability to execute deployment commands (wrangler), perform AI inference (env.AI.run), and automate browsers (puppeteer).
Sanitization: Although the documentation text advises users to validate URLs and sanitize content, the provided examples interpolate untrusted data directly into the LLM context.
[COMMAND_EXECUTION] (SAFE): Execution of local CLI tools for deployment purposes.
Evidence: The scripts/cloudflare_deploy.py script uses subprocess.run to invoke the wrangler CLI for worker deployment.
Context: This is the primary intended function of the DevOps skill. The script follows security best practices by using list-based arguments to prevent shell injection vulnerabilities.

devops