The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill is designed to read, parse, and synchronize content from .env files across different git worktrees. These files are standard locations for storing sensitive data such as API keys, tokens, and database credentials. Processing and copying these secrets increases the surface area for potential exposure or misuse. Evidence: Step 5 instructions to parse and copy .env keys and values.\n- [COMMAND_EXECUTION]: The skill executes installation commands for various package managers (npm, yarn, pnpm, bun). These operations can run arbitrary code through package lifecycle scripts, which poses a risk if the dependencies in the worktrees have not been thoroughly audited. Evidence: Step 4 instructions to run install commands based on detected lockfiles.\n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface as it ingests untrusted data from local .env files and git command output without using boundary markers or sanitization. This surface is combined with high-impact capabilities such as file-writing and command execution. Ingestion points: .env files (SKILL.md) and git worktree porcelain output (SKILL.md). Boundary markers: Absent. Capability inventory: git rebase, npm/yarn/pnpm/bun install, and .env file writes. Sanitization: Absent.

lgnh-rebase-worktrees