truto-integrations-build

CRITICAL E005: Suspicious download URL detected in skill instructions.

• Suspicious download URL detected (high risk: 0.80). The list includes direct links to an installer PowerShell script and a shell script (https://cli.truto.one/install.ps1 and https://cli.truto.one/install.sh), which are high-risk because piping remote scripts to a shell/PowerShell can execute arbitrary/malicious code; the remaining URLs are documentation or API/spec endpoints and are low risk.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). SKILL.md describes runtime ingestion of outsider-authored content from user-supplied source URLs (openapi/postman/graphql/doc pages) via the agent’s web_search/web_fetch and doc scraping tools, which feed that fetched free text into the LLM context during Phase A (autonomous build) and the docs phase.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). I flagged runtime-download-and-execute install scripts (https://cli.truto.one/install.sh and https://cli.truto.one/install.ps1) because they are fetched and executed during setup, and I flagged example source URLs (e.g. https://api.acme.com/openapi.json, https://docs.crisp.chat/guides/rest-api/, https://docs.crisp.chat/static/data/collections/rest-api-v1.postman) because the CLI fetches those docs at build time and the fetched content is injected into the agent's prompts/system context to drive the build.