upgrade-assessment
Pass
Audited by Gen Agent Trust Hub on Mar 26, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses bash commands to detect the environment and scan the codebase for specific patterns. These commands are executed locally to inspect files like
CLAUDE.md,hooks.json, and source code directories. Evidence:rules/detection-checks.mdandrules/codebase-scan-patterns.md.- [PROMPT_INJECTION]: The skill has a surface for indirect prompt injection via the research workflow in Phase 2. It fetches external documentation and changelogs from the web usingWebFetch. If an attacker were to poison these documentation sources with malicious instructions, the agent might attempt to follow them.\n - Ingestion points: External content fetched via
WebFetchin Phase 2 (SKILL.md).\n - Boundary markers: None. The skill does not instruct the agent to ignore instructions within the fetched content.\n
- Capability inventory:
Bash,Task,WebFetch,WebSearch,Readtools are allowed (SKILL.md).\n - Sanitization: No sanitization or validation is performed on the retrieved web content.
Audit Metadata