Security risk report for @forgecat/vercel-labs_agent-skills_vercel-cli-with-tokens v0.0.9
Source Integrity
Low
Profile is sourced from upstream Vercel Labs agent-skills repository (pinned commit 063bee94c3f4df8453406c830b0a7df0f2860278), with clear provenance documentation in LICENSE-SOURCE.md and upstream-README.md.
ForgeCat has transparently labeled its own additions (README.md, MIT-STANDARD-REFERENCE.md) separately from upstream content, preserving the original skill boundary and avoiding injection of duplicate global rules.
Agent Intent
Low
Content describes legitimate Vercel CLI workflows with token-based authentication; no instructions to ignore system prompts, exfiltrate credentials to external hosts, or hide its own instructions.
Guidance on token handling (Step 1) is defensive security practice—explicitly warns against exposing tokens in shell history and recommends environment variable export, which is the secure pattern.
No evidence of prompt injection, role hijacking, guidance poisoning, or system prompt leakage; the skill documents standard Vercel deployment and environment management procedures.
Details
Evidence
"do not pass it as a `--token` flag. Putting secrets in command-line arguments exposes them in shell history and process listings."
No instructions to read ~/.ssh, ~/.aws, or exfiltrate data; no directives to disable security defaults or install backdoored dependencies.
Permissions
Low
No tool definitions or MCP servers are declared in this profile; it is purely instructional documentation for the Vercel CLI.
Authority is scoped to Vercel CLI operations (deploy, env management, domain management, inspection) and aligns with the stated purpose of token-based Vercel project management.
No shell execution, file write, file delete, or subagent permissions are requested; the skill assumes the agent has access to the Vercel CLI binary and local filesystem for reading .env and .vercel/ directories.
Details
Evidence
No declared tools or MCP servers in the profile.
Instructions are procedural guidance for CLI commands, not permission grants.
MCP Risk
Low
No MCP servers are defined in this profile; it is a documentation-only skill.
Content does not reference external binaries, arbitrary code execution, or unrestricted network/filesystem access beyond standard Vercel CLI and local .env/.vercel/ directory operations.
Details
Evidence
MCP servers section is empty: "(none)"
No hidden instructions in tool descriptions; no unknown binary execution.
ForgeCat