explain-diff
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data including git diffs, branch content, and pull request metadata which may contain malicious instructions designed to influence the agent during the research and build phases.
- Evidence Chain:
- Ingestion points: Git diffs, PR metadata, and source code are ingested in SKILL.md Steps 1 and 2.
- Boundary markers: The skill contains instructions to ignore unproven motivation and remove unsupported claims in SKILL.md Step 3 and REFERENCE.md.
- Capability inventory: The agent performs file system read/write operations, executes git commands, and uses a browser tool to execute generated code as described in SKILL.md Step 4.
- Sanitization: REFERENCE.md requires escaping source text, quiz options, and feedback before insertion into HTML.
- [DYNAMIC_EXECUTION]: In SKILL.md Step 4, the agent is instructed to use a browser tool to load the generated HTML file and test the inline JavaScript quiz logic. While the JS is generated from templates, the input data could theoretically influence the executable content.
- [COMMAND_EXECUTION]: The skill uses shell commands to interact with git (e.g., fetching diffs, resolving branches) and handles user-provided PR URLs.
- [DATA_EXPOSURE]: The skill reads source code, including definitions, callers, contracts, and tests, to extract information for the generated report.
Audit Metadata