normalize-agent-instructions

Pass

Audited by Gen Agent Trust Hub on Jul 26, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill utilizes standard command-line utilities (ripgrep, git) for file discovery and management. These commands are used in a controlled manner with static flags to limit scope to specific filenames and exclude sensitive directories like .git and node_modules.\n- [COMMAND_EXECUTION]: Executes local shell commands (rg, git) to scan the project and manage file migrations. The execution is gated by internal checks and is designed to be non-destructive by default.\n- [SAFE]: There are no network requests or external data transfers. The skill's operations are strictly local to the user's project environment.\n- [PROMPT_INJECTION]: The skill processes the contents of external instruction files, which presents a surface for indirect prompt injection. However, the risk is mitigated by the skill's narrow focus on file movement and standardization rather than instruction interpretation.\n
  • Ingestion points: Reads metadata and content from AGENTS.md and CLAUDE.md files during project scanning.\n
  • Boundary markers: None. The content is read directly from the filesystem without specific delimiters.\n
  • Capability inventory: Execute shell commands (rg, git), perform file moves (git mv), and write files.\n
  • Sanitization: None. The content is moved or rewritten to maintain its original form without filtering.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 26, 2026, 07:58 AM
Security Audit — agent-trust-hub — normalize-agent-instructions