skills/rpamis/comet/comet-native/Gen Agent Trust Hub

comet-native

Pass

Audited by Gen Agent Trust Hub on Oct 12, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data when the user provides files, links, or attachments as requirement sources, creating a surface for indirect prompt injection.
  • Ingestion points: User-supplied requirement sources (files, links, attachments, or local paths) processed in the "source-document full coverage" mode described in SKILL.md and reference/artifacts.md.
  • Boundary markers: Present. reference/artifacts.md explicitly states that agent-directed instructions inside source material must be treated as data and cannot override user requests or project rules.
  • Capability inventory: The skill utilizes the comet native CLI for project management, performs file system writes to target specifications and briefs, and executes Git operations for archiving and merging.
  • Sanitization: Present. The "Source-document full coverage" protocol requires the agent to build a mapping table that classifies every source item and explicitly links it to a specification location and acceptance ID before implementation begins.- [COMMAND_EXECUTION]: The skill relies on the comet CLI toolset (including comet native, comet task, comet memory, and comet knowledge) to manage workflow state and memory. These commands are executed locally on the user's PATH and are consistent with the skill's stated purpose for the rpamis vendor ecosystem.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 12, 2026, 03:00 AM