Feature request: list Correctover Security Hook as a curated plugin (PreToolUse secret/SSRF gate)

[DSHCorrectover]

Context

Kimi Code loads plugins from plugins/marketplace.json (official + curated tiers). Curated precedents already include community security/tooling plugins (e.g. obra/superpowers, GoogleChrome/modern-web-guidance). This is a request to list a community security plugin.

What the plugin does

A PreToolUse security gate for Kimi Code sessions:

• Hook 1 (WriteFile|StrReplaceFile): scans file content the agent is about to write for exposed cloud credentials, API keys, tokens, and secret-like material — blocks the write if a critical finding is returned.
• Hook 2 (Shell): scans shell commands for SSRF/credential-exfiltration patterns (e.g. piping env vars/credentials to network endpoints) — blocks dangerous commands before execution.

The hook itself is a thin, self-contained Node.js wrapper (hooks/correctover-hook.mjs, zero external dependencies). It delegates scanning to npx correctover-scan, an existing MCP config/code security scanner CLI. Findings are returned to the agent via hook stderr as structured block reasons.

Plugin structure (already in official manifest format)

• kimi.plugin.json — manifest with two [[hooks]] PreToolUse entries (event, matcher, command, timeout), aligned with the existing hook config shape used by Kimi Code
• hooks/correctover-hook.mjs — the gate (MIT-licensed)
• selftest/ — 8 local test scenarios (clean write passes; credential leak in a written file is blocked; leaky shell command is blocked; malicious MCP config detected), runnable with bash selftest/run-hook-selftest.sh
• README.md — install, environment variables (CORRECTOVER_SCAN_BIN, CORRECTOVER_HOOK_BLOCK_LEVEL), limits (free tier 50 scans/day), honest coverage statement

Proposed marketplace.json entry

{
  "id": "correctover-security-hook",
  "tier": "curated",
  "displayName": "Correctover Security Hook",
  "description": "PreToolUse gate: blocks secret exposure in written files and SSRF/credential-exfil shell commands before they execute, via correctover-scan.",
  "homepage": "https://github.com/DSHCorrectover/kimi-code-security-hook",
  "keywords": ["security", "hooks", "pretooluse", "secret-scanning", "ssrf", "mcp"],
  "source": "https://github.com/DSHCorrectover/kimi-code-security-hook"
}

Verification / next steps

• The plugin repo is public, MIT-licensed, and self-test passes locally (clean code → exit 0 allow; seeded credential leak → exit 2 block with Chinese-language block reason fed back to the agent).
• The manifest targets the documented plugin hook schema; if any field differs from the current kimi.plugin.schema.json, happy to adjust — I couldn't fetch the schema URL directly (JS-rendered).
• Per CONTRIBUTING.md: happy to open a PR against plugins/marketplace.json after a maintainer /approves this issue, or adjust any review feedback first.

Thanks for Kimi Code — happy to move fast on any requested changes.