product-help-center

Pass

Audited by Gen Agent Trust Hub on Oct 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill represents a vulnerability surface as it is designed to ingest and process data from external sources (defined in data/sources.json) and user-provided requirements to generate AI-driven documentation systems.
  • Ingestion points: Data enters the agent's context through the sources.json catalog and user-supplied design requirements in SKILL.md.
  • Boundary markers: The skill explicitly mandates a 'Retrieval-first' policy in references/ai-integration.md, requiring citations for factual claims and stating that assistants should answer only from retrieved evidence.
  • Capability inventory: The skill utilizes local scripts for operations, including append_learning.py for persistence of learnings and validate_sources.py for network-based validation of sources.
  • Sanitization: No explicit sanitization logic is provided for the internal feedback scripts, though they are scoped to local file operations.
  • [EXTERNAL_DOWNLOADS]: The data/sources.json file contains a curated list of external URLs for official documentation and protocol specifications.
  • Evidence: The skill references domains such as support.zendesk.com, intercom.com, docs.readme.com, gitbook.com, mintlify.com, developers.google.com, learn.microsoft.com, and docs.stripe.com.
  • Assessment: All referenced external sources are official documentation portals for well-known and established technology services and cloud providers, which are treated as safe under security protocols.
  • [COMMAND_EXECUTION]: The skill includes instructions and scripts for performing maintenance and operational tasks.
  • Evidence: SKILL.md instructs the agent to use agents-skills-feedback-loop/scripts/append_learning.py to record new patterns, and scripts/validate_sources.py provides a utility to validate the integrity of the source catalog using Python's standard urllib library.
  • Assessment: These are standard operational tools for maintaining skill state and data integrity, executing locally within the agent's environment without processing untrusted shell input.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 6, 2026, 02:02 PM