skills/phrazzld/claude-config/backlog/Gen Agent Trust Hub

backlog

Pass

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: Risk of indirect prompt injection via external GitHub issue content.
  • Ingestion points: The skill fetches issue titles, labels, and bodies from GitHub using the gh issue list command in SKILL.md.
  • Boundary markers: No delimiters or explicit instructions are used to separate issue content from agent instructions.
  • Capability inventory: The skill computes statistics, identifies staleness, and invokes the /issue lint command.
  • Sanitization: The fetched data is processed directly for readiness scoring and reports without validation or filtering of potential embedded instructions.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 27, 2026, 02:26 PM