ollygarden-cli
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data from telemetry and scan reports that may contain adversarial instructions. Ingestion points: JSON output from ollygarden CLI commands, including findings, insights, and service metadata as documented in SKILL.md and references/commands.md. Boundary markers: Present; the skill provides strict instructions to treat all API-derived fields as untrusted data, not as instructions for the agent. Capability inventory: The skill utilizes the ollygarden CLI for network-based queries, git for local repository correlation, and jq for data processing. Sanitization: Present; instructions mandate UUID shape validation for all identifiers, mandatory redaction of apparent secrets, and the use of bounded sanitized summaries for malicious payloads.
- [COMMAND_EXECUTION]: The skill relies on shell execution of the ollygarden CLI binary and git tools. It implements robust error handling by defining specific actions for different CLI exit codes (0-7) and requires explicit authorization gates for mutating operations such as authentication changes or webhook modifications.
Audit Metadata