tiangong-kb-report-search

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The skill explicitly allows/asks for api_key/report_api_key to be passed as JSON fields which are forwarded as CLI flags (and take precedence over env), meaning the agent would need to include secret values verbatim in generated commands/flags.

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

• Third-party content exposure detected (medium risk: 0.30). The workflow path takes outsider-authored JSON input (including arbitrary query text or full request_file content) in report_search.sh, then passes it to npx @tiangong-ai/cli@latest research search --sources report --json via --query or --input, meaning the LLM ingests user-supplied free text at runtime.

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

• Potentially malicious external URL detected (high risk: 1.00). The wrapper defaults to running "npx @tiangong-ai/cli@latest" at runtime (seen in scripts/report_search.sh and SKILL.md), which fetches and executes remote npm package code that the skill relies on to perform searches.