unfold
Pass
Audited by Gen Agent Trust Hub on Jul 25, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructions in
references/triage.md,references/change.md, andreferences/review.mddirect the agent to run shell commands to reproduce issues, execute test suites, and verify code changes. This capability is inherent to its primary purpose of aiding software development and maintenance. - [PROMPT_INJECTION]: The skill has an indirect prompt injection surface due to its processing of repository source code and local artifacts.
- Ingestion points: Source code files and
.unfold/directory artifacts (e.g.,manifest.json,architecture.md) are read by the agent as described inSKILL.md. - Boundary markers: The protocol utilizes specific evidence labels (
E,I,?) and YAML frontmatter in its artifacts (e.g., inreferences/artifacts.md) to structure and delimit information. - Capability inventory: The agent can perform file system reads and execute shell commands for testing and diagnostics, as defined in the mode-specific references.
- Sanitization: The provided instructions do not specify explicit sanitization or filtering for the content of the analyzed files before they are processed by the agent.
Audit Metadata