craft-telegram-bot-surface
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides instructions for building interfaces that ingest and process untrusted external data (user messages and callback data), creating a potential attack surface.
- Ingestion points: The skill references several points where untrusted data enters the agent's context, including
callback_querydata (referenced inreferences/callback-protocol.md), andForceReplyor free-text inputs (referenced inreferences/interaction-primitives.md). - Boundary markers: The skill mentions security boundaries, such as the instruction to "Escape Markdown punctuation in dynamic user/domain strings before interpolating them" in
references/rich-message-format.md, though these are not absolute protections against adversarial input. - Capability inventory: The skill utilizes capabilities like
sendRichMessage,editMessageText, andanswerCallbackQueryto interact with users and modify message content based on input. - Sanitization: The skill recommends several sanitization and validation practices, including server-side authorization checks, idempotent handlers, and Markdown escaping to mitigate risks associated with processing external data.
Audit Metadata