diff-summary-quiz
Pass
Audited by Gen Agent Trust Hub on Sep 10, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes local Python scripts (
collect_diff_evidence.pyandgenerate_summary_report.py) and system binaries (git,gh) to collect repository data and generate reports. - Evidence:
- Subprocess calls are managed with
shell=Falseand isolated Python mode (-I). - A hardened Git configuration is enforced, including flags to disable promisor helpers, replace objects, and external diff/textconv drivers (
--no-ext-diff,--no-textconv). - Execution environment is sanitized by removing sensitive variables and using a restricted
PATHresolved via absolute file paths. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from git diffs, PR metadata, and untracked repository files, which could contain malicious instructions.
- Ingestion points:
scripts/collect_diff_evidence.pycaptures output fromgit diff,gh pr view, and reads contents of untracked files. - Boundary markers:
references/diff-summary-workflow.mdprovides explicit instructions to the agent to treat all evidence as inert data and ignore any instructions or links embedded within it. - Capability inventory: The skill can write files to the
.diff-summaries/directory and execute shell commands through the collector script. - Sanitization: The
generate_summary_report.pyscript applies HTML escaping to content embedded in the report. The collector script implements a 'clean-filter preflight' to prevent Git from executing potentially malicious filter drivers defined in the repository. - [DYNAMIC_EXECUTION]: The skill generates interactive HTML reports containing JavaScript for UI functionality and attempts to launch the system's default browser to display them.
- Evidence:
scripts/generate_summary_report.pyusesos.startfile(Windows) orxdg-open/open(Linux/macOS) with a sanitized environment to open the generated HTML files.- The
assets/summary-template.htmlfile includes significant JavaScript for managing user interactions, comprehension quizzes, and local storage of comments.
Audit Metadata