fixing-harness

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it reads and processes user-controlled configuration files (CLAUDE.md, AGENTS.md, and project rules) as part of its auditing workflow.
  • Ingestion points: The workflow in SKILL.md (Step 1) involves reading all files related to the agentic harness from the target project directory.
  • Boundary markers: The skill does not implement delimiters or explicit instructions for the agent to ignore or isolate natural language directives contained within the files being audited.
  • Capability inventory: The agent is granted capabilities to perform file system modifications and execute shell commands (wc, grep, git, find).
  • Sanitization: No sanitization or validation of the content within the harness files is performed prior to the agent processing them for refactoring.
  • [COMMAND_EXECUTION]: The skill utilizes several standard shell utilities to perform project analysis and verify the effectiveness of the refactoring.
  • Evidence: The instructions call for the use of tokei, find, ls, wc -l, git ls-files, grep, and git mv to inventory files and validate rule counts or path scopes.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 01:57 AM
Security Audit — agent-trust-hub — fixing-harness