feedback-triage

Pass

Audited by Gen Agent Trust Hub on Oct 15, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted reader feedback sourced from GitHub issues and local files.
  • Ingestion points: Untrusted data enters the context via gh issue list (SKILL.md Step 2.1) and by reading reader feedback files (feedback/round-{N}/{reader-kebab}.md).
  • Boundary markers: The skill instructs the agent to record notes using a specific template (references/feedback-template.md) and includes explicit instructions to treat reader input as material to be weighed, not instructions to be followed.
  • Capability inventory: The skill can execute shell commands (git, gh, story CLI) and write files (feedback records and synthesis reports).
  • Sanitization: The instructions require validating input (labels, usernames) and checking for hidden characters or HTML comments in external notes before processing.
  • [COMMAND_EXECUTION]: The skill performs several shell operations using git (tagging, committing), gh (fetching issues), and a custom story CLI tool.
  • The skill includes defensive instructions to validate labels (e.g., checking for specific patterns like ch03-p12) and usernames before they are used as command arguments or filenames to mitigate injection risks.
  • [EXTERNAL_DOWNLOADS]: The skill refers to and suggests using templates hosted on the author's GitHub repository (github.com/danjdewhurst/story-skills).
  • These templates are used for GitHub Actions workflows and issue forms.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 15, 2026, 03:10 PM