a1-message
Pass
Audited by Gen Agent Trust Hub on Aug 30, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted data such as correspondence and notes. This creates a surface for indirect prompt injection.\n
- Ingestion points: The skill ingests user-supplied facts, notes, correspondence, transcripts, and repository-level context files as specified in
SKILL.mdandreferences/source-policy.md.\n - Boundary markers: The
source-policy.mdfile defines strict rules against merging conflicting instructions and requires the agent to identify and stop if the communication job is out of scope.\n - Capability inventory: The skill is restricted to text generation and does not have tools or permissions for command execution, network access, or unauthorized file system operations.\n
- Sanitization: The
references/final-qa.mdchecklist provides mandatory validation steps to ensure all output is source-faithful and that no fabricated details are included from the input data.\n- [EXTERNAL_DOWNLOADS]: The skill references a support link (https://t.me/a1_marketing_skills) inSKILL.mdandreferences/final-qa.md. This is a static URL provided for user communication and feedback and is not used for automated code retrieval or script execution.
Audit Metadata