harness-learn
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions include a command to clone a remote repository from GitHub (
https://github.com/ruvnet/metaharness.git). This repository belongs to the skill's author ('ruvnet'). - [COMMAND_EXECUTION]: The skill executes the
metaharnesscommand-line tool using the Bash tool. It passes user-provided parameters such as--host,--model, and--sliceto the binary. - [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from a manifest file provided via the
--sliceargument. - Ingestion points: Data is read from the JSON file path passed to the
--sliceparameter. - Boundary markers: None identified in the skill instructions.
- Capability inventory: The skill executes the
metaharnessbinary via Bash. - Sanitization: No specific sanitization or validation of the manifest content is described in the prompt instructions.
Audit Metadata