rules-from-coding-agent-failures

Pass

Audited by Gen Agent Trust Hub on Jun 26, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill possesses an Indirect Prompt Injection surface by ingesting and processing failure logs to generate system-level rules.
  • Ingestion points: The promote workflow reads failure descriptions from files in 'docs/reflection-log/' which are considered untrusted input.
  • Boundary markers: The skill implements a 'W1 floor' requiring at least three separate entries before action and mandates explicit user confirmation for all changes.
  • Capability inventory: The skill can modify 'AGENTS.md' and create new executable hooks or CI gates.
  • Sanitization: The workflow relies on manual human review of proposed instructions rather than automated sanitization of log content.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 26, 2026, 02:21 AM
Security Audit — agent-trust-hub — rules-from-coding-agent-failures