aidoc-triage

Pass

Audited by Gen Agent Trust Hub on Nov 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on shell commands including gh, git, grep, and sed to interact with the GitHub API and local repository files. These commands are used to fetch issue data, inspect the local repository state, and modify issue metadata.
  • [INDIRECT_PROMPT_INJECTION]: The skill fetches and processes content from GitHub issues (titles, bodies, and comments), which is untrusted external data.
  • Ingestion points: Untrusted data is ingested via gh issue view in the "Fetch the issue live" section of SKILL.md.
  • Boundary markers: The instructions do not define boundary markers to encapsulate issue content or provide instructions to ignore embedded commands.
  • Capability inventory: The agent can perform write operations such as labeling (gh issue edit), creating labels (gh label create), transferring issues (gh issue transfer), and local file system interactions.
  • Sanitization: There is no mention of escaping or sanitizing the issue content before the agent processes it or incorporates it into its reasoning. The instruction to treat thread corrections as superseding the body increases the susceptibility to malicious instructions embedded in comments.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 8, 2026, 11:07 PM