docs-workflow
Fail
Audited by Gen Agent Trust Hub on Feb 15, 2026
Risk Level: HIGHCREDENTIALS_UNSAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- Data Exposure (HIGH): The skill maintenance process reads sensitive local files including .env and .env.example to extract environment variable names. Evidence: commands/docs-claude.md specifies reading envVars from .env using parseEnvFile().
- Indirect Prompt Injection (HIGH): The skill ingests untrusted content from project files and has permission to modify the filesystem and execute system commands, creating a significant injection surface. 1. Ingestion points: package.json, CLAUDE.md, README.md, .env, wrangler.jsonc. 2. Boundary markers: Absent. 3. Capability inventory: File system write access, git log, npm view. 4. Sanitization: Absent.
- Command Execution (LOW): The skill executes read-only system commands for auditing project state. Evidence: commands/docs-update.md uses git log for date comparisons and npm view for version verification via the npm registry.
Recommendations
- AI detected serious security threats
Audit Metadata