self-improvement-ci

Pass

Audited by Gen Agent Trust Hub on Apr 22, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Installs the gh-aw extension from the official github organization on GitHub.
  • [COMMAND_EXECUTION]: Executes GitHub CLI (gh) commands for authentication, extension management, and workflow execution.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted data from pull request checks and workflow logs. Ingestion points include PR-scoped check results and Action outcomes. Boundary markers are not explicitly defined in the instructions. Capability inventory includes emitting structured learning candidates and potentially updating repository context files (e.g., CLAUDE.md) via the gh aw run --push command. Sanitization is addressed via a recurrence policy requiring patterns to appear in at least three separate runs and two distinct tasks within 30 days before being recommended for promotion.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 22, 2026, 12:10 AM