dev-loop-pr-writer

Pass

Audited by Gen Agent Trust Hub on Jul 8, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill contains Python code (PRDescriptionGenerator class in SKILL.md) and a shell script (scripts/generate-pr-body.sh) that utilize subprocess.run and direct shell execution to run git commands (git diff, git log, git stat). While standard for development tools, these represent a capability to execute shell commands within the environment.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it processes external, untrusted data including git commit messages, git diff outputs, and issue tracker references to generate PR summaries.
  • Ingestion points: Commit messages and issue descriptions enter the agent context through the workflow context and generation scripts.
  • Capability inventory: The skill possesses shell execution capabilities through the use of subprocess.run and shell scripts.
  • Boundary markers: The skill lacks explicit delimiters or instructions to prevent the agent from obeying malicious commands embedded within the processed git metadata.
  • Sanitization: No validation or escaping logic is provided to sanitize input data (e.g., commit messages) before it is interpolated into PR descriptions or prompts.
  • [DATA_EXPOSURE]: The provided scripts access the repository's internal git history and diff stats. While functional, this exposes sensitive metadata regarding the project's development history to the agent.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 8, 2026, 12:22 PM
Security Audit — agent-trust-hub — dev-loop-pr-writer