harness-verify-before-move

Pass

Audited by Gen Agent Trust Hub on Jun 14, 2026

Risk Level: SAFECOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill utilizes common development utilities such as git, npm, pip, gh, grep, and jq to verify code status and synchronize project data. These operations are essential to the skill's stated purpose of managing phase transitions.
  • [REMOTE_CODE_EXECUTION]: Step 2 dynamically identifies and executes build, lint, and test commands defined in project configuration files such as package.json, Makefile, and pyproject.toml. The skill includes a safety check to ask the user for confirmation if the commands are not recognized.
  • [PROMPT_INJECTION]: The skill processes project documentation and status files which could contain instructions intended to influence the agent. However, the skill maintains a rigid 10-step execution sequence and requires user confirmation for impactful actions.
  • Ingestion points: Processes content from .harness/phases/, implementation-notes.md, MEMORY.md, and the root README.md.
  • Boundary markers: No specific delimiters are used to isolate untrusted data during processing.
  • Capability inventory: Includes subprocess execution of build tools, file system modifications (archiving/deleting), and GitHub API interactions.
  • Sanitization: No explicit sanitization or filtering of the ingested document content is described.
  • [EXTERNAL_DOWNLOADS]: Step 7 interacts with GitHub's official API using the gh CLI to sync milestones and project boards. These connections are restricted to the repository owner and name specified in the project's local configuration.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 14, 2026, 10:13 PM
Security Audit — agent-trust-hub — harness-verify-before-move