draft-release-notes
Pass
Audited by Gen Agent Trust Hub on Oct 3, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses a Python helper script
scripts/collect_release_evidence.pyto execute Git commands such asgit rev-parse,git ls-tree,git show, andgit log. These operations are performed usingsubprocess.runwith list-based arguments (shell=False), which is the secure pattern for executing external commands. The commands are scoped to the repository directory and are used solely to gather evidence for release notes. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository's Git commit history and file content, which could theoretically contain malicious instructions targeting the agent.
- Ingestion points: The helper script
scripts/collect_release_evidence.pyreads commit subjects and file content from the repository via Git. - Boundary markers: The instructions do not define specific delimiters for the gathered evidence, but provide a structured workflow for the agent to follow.
- Capability inventory: The agent is granted the capability to update documentation files in
docs/about-nemo-relay/release-notes/. - Sanitization: The script performs classification of commits using regex, and the skill instructions explicitly require the agent to verify each candidate claim against the source code and public documentation before including it in the draft.
Audit Metadata