logic-fix-all

Pass

Audited by Gen Agent Trust Hub on Jul 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: Executes shell commands to audit the repository, including git ls-files, find, git rev-parse, git diff, and git checkout.
  • [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its autonomous pipeline design. It ingests untrusted data from all repository files (source code, docs, config) to generate code remedies.
  • Ingestion points: Phase 1c identifies source code, runtime configurations, constraint files, and behavioral documentation as scan targets.
  • Boundary markers: The process lacks explicit boundary markers or instructions to ignore embedded commands within the processed files.
  • Capability inventory: The skill has the capability to execute shell commands and autonomously modify files on the local filesystem.
  • Sanitization: No sanitization or filtering of external file content is described before the content is used in code-generation prompts.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 5, 2026, 09:37 AM
Security Audit — agent-trust-hub — logic-fix-all