seed-test-data
Pass
Audited by Gen Agent Trust Hub on Nov 10, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it is designed to ingest and analyze untrusted data from the local project environment.
- Ingestion points: In
SKILL.md(Protocol steps 2, 4, 5) andreferences/seed-test-data-skill-review.md(R1), the skill instructions require the agent to read project configuration files (docs/project-config.json), reference documents (seed-test-data-reference.md), and existing source code/seeders. - Boundary markers: While the skill provides instructions to redact sensitive data like credentials or PII, it lacks explicit boundary markers or instructions to ignore embedded natural language instructions within the code it analyzes.
- Capability inventory: The skill possesses the capability to write files (Implementation step), spawn sub-agents (
spawn_agent), and execute local scripts. - Sanitization: There is no evidence of sanitization or filtering for natural language instructions that might be embedded in code comments or configuration values within the ingested files.
- [COMMAND_EXECUTION]: The skill's 'Sub-Agent Routing' section in
SKILL.mdincludes instructions to execute a local Python script:python .claude/scripts/code_graph trace <file> --direction both --json. This is used to analyze the codebase blast radius, which is a legitimate part of its functionality but involves the execution of shell commands.
Audit Metadata