ln-52-delivery-reviewer
Pass
Audited by Gen Agent Trust Hub on Oct 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted data from repositories (code, diffs, and manifests) which serves as a primary vector for indirect prompt injection.
- Ingestion points: Repository files, diffs, documentation, and configuration manifests (SKILL.md sections 1.5, 2.2, 4.7).
- Boundary markers: The instructions do not define specific delimiters or boundary markers to isolate untrusted repository content from the agent's logic.
- Capability inventory: The skill authorizes the execution of repository-defined commands such as build, lint, and test scripts (SKILL.md section 4.7).
- Sanitization: No explicit sanitization or validation of the ingested repository content or discovered commands is specified before processing or execution.
- [COMMAND_EXECUTION]: The skill instructs the agent to discover and run commands defined within the repository being reviewed, which could lead to the execution of malicious scripts if the repository is compromised.
- Evidence: Checklist section 4 explicitly directs the agent to "Discover commands from repository docs, tool configuration, and manifests" and to "Run narrow checks first, then required build, lint, type, test, migration, and smoke gates."
- Impact: Executing arbitrary scripts from an untrusted repository (e.g., scripts defined in a
package.jsonorMakefile) allows for potential remote code execution on the host machine.
Audit Metadata