alan-test-feature
Pass
Audited by Gen Agent Trust Hub on Nov 13, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill discovers and executes development server commands (e.g.,
npm run dev,docker compose up) found within project configuration files and uses theagent-browserCLI for automated interactions. - [EXTERNAL_DOWNLOADS]: Installs the
agent-browserutility from Vercel Labs using the NPM package manager if the tool is not found in the execution environment. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted repository content, such as git diffs and configuration files, to determine the testing strategy and commands to execute.
- Ingestion points: Analyzes project files including
package.json,git diff,git log,docker-compose.yml, and framework configurations to derive execution logic. - Boundary markers: Absent; instructions do not specify delimiters for distinguishing repo data from agent instructions.
- Capability inventory: Arbitrary shell execution via bash, network access for artifact uploads via
curl, and file system writes restricted to/tmp/alan-captures/. - Sanitization: Absent; the agent is instructed to trust the contents of the repository's build scripts and configuration for server discovery.
Audit Metadata