post-build-flow
Pass
Audited by Gen Agent Trust Hub on Jul 13, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill contains instructions to suppress conversational responses (e.g., 'do not acknowledge first', 'do not write a message first') when specific workflow events occur. This is a user interface optimization intended to favor tool-provided inline cards over redundant chat messages.
- [PROMPT_INJECTION]: The skill describes an attack surface for indirect prompt injection via external workflow trigger data. 1. Ingestion points: External data is ingested as inputData for Form, Webhook, and Chat triggers as documented in references/trigger-input-data-shapes.md. 2. Boundary markers: No specific delimiters or 'ignore' instructions are present to isolate external data from agent context. 3. Capability inventory: The skill utilizes the verify-built-workflow and executions(action='run') tools in SKILL.md which process trigger payloads. 4. Sanitization: No sanitization, escaping, or validation logic is defined for the trigger input data within the skill.
Audit Metadata