a1-presentation-copy

Pass

Audited by Gen Agent Trust Hub on Aug 24, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill provides a legitimate utility for presentation drafting with no indicators of malicious intent or hidden executable code.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied research and strategy materials, which creates a potential surface for indirect prompt injection. However, the skill implements significant mitigations.
  • Ingestion points: External materials provided by the user (research, data, strategy) are ingested into the agent context via standard file reads or prompt input.
  • Boundary markers: The SKILL.md includes an 'Entry Contract' that defines specific scope boundaries (e.g., only new presentations, no .pptx production) and mandates stopping if a request is out of scope.
  • Capability inventory: The skill's capabilities are restricted to text-based generation of outlines and slide copy; it does not utilize tools for file system modification, command execution, or network communication.
  • Sanitization: The skill follows a 'Source Policy' (references/source-policy.md) which requires all takeaways and conclusions to be strictly supported by supplied source material, effectively preventing the agent from following unintended instructions embedded in the source data.
  • [DATA_EXPOSURE_AND_EXFILTRATION]: The skill reads project-specific marketing context from standard repository locations (.agents/marketing-context.md, .claude/marketing-context.md). These files are intended for project-level context and do not involve sensitive system credentials or environment variables. The skill also includes a legitimate support footer linking to a Telegram channel (t.me/a1_marketing_skills), which is a disclosed vendor resource for user assistance.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 24, 2026, 07:29 PM
Security Audit — agent-trust-hub — a1-presentation-copy