Kjell Tore Guttormsen f418a8fe08 feat(llm-security-copilot): port llm-security v5.1.0 to GitHub Copilot CLI

Full port of llm-security plugin for internal use on Windows with GitHub
Copilot CLI. Protocol translation layer (copilot-hook-runner.mjs)
normalizes Copilot camelCase I/O to Claude Code snake_case format — all
original hook scripts run unmodified.

- 8 hooks with protocol translation (stdin/stdout/exit code)
- 18 SKILL.md skills (Agent Skills Open Standard)
- 6 .agent.md agent definitions
- 20 scanners + 14 scanner lib modules (unchanged)
- 14 knowledge files (unchanged)
- 39 test files including copilot-port-verify.mjs (17 tests)
- Windows-ready: node:path, os.tmpdir(), process.execPath, no bash

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-04-09 21:56:10 +02:00

2 KiB

Raw Blame History

name	description
security-plugin-audit	Audit a plugin for security risks, permission analysis, and trust assessment before installation

Plugin Audit

Dedicated plugin security audit with trust verdict. Accepts local paths or repository URLs.

Step 1: Resolve Target

If $ARGUMENTS starts with https:// → clone the repo:

node <plugin-root>/scanners/lib/git-clone.mjs clone "<url>" [--branch <branch>]

Set clone_path and target.

Otherwise → target = $ARGUMENTS or current directory

IMPORTANT: Cleanup Guarantee (remote)

If cloned, cleanup MUST run regardless of outcome:

node <plugin-root>/scanners/lib/git-clone.mjs cleanup "<clone_path>"

Step 2: Pre-extraction (remote only)

node <plugin-root>/scanners/lib/fs-utils.mjs tmppath "plugin-extract.json"
node <plugin-root>/scanners/content-extractor.mjs "<target>" --output-file "<evidence_file>"

Step 3: Inventory

Read plugin manifest (plugin.json). Glob for all components: commands, agents, hooks, skills, MCP configs. Build permission matrix: what tools each component can access.

Step 4: Security Analysis

Read <plugin-root>/knowledge/skill-threat-patterns.md and <plugin-root>/knowledge/secrets-patterns.md.

Analyze all files for 7 threat categories. If remote, analyze from evidence package (never read untrusted files directly). Check hook registration vs scripts (ghost hooks). Cross-reference description vs tools permissions.

Step 5: Report

Output: Plugin metadata, component inventory, permission matrix, hook analysis, all findings by severity.

Trust Verdict:

Install — No critical or high findings, permissions appropriate
Review — High findings present but potentially justified, manual review recommended
Do Not Install — Critical findings, active exfiltration, or injection detected

Step 6: Cleanup

node <plugin-root>/scanners/lib/git-clone.mjs cleanup "<clone_path>"
node <plugin-root>/scanners/lib/fs-utils.mjs cleanup "<evidence_file>"

2 KiB Raw Blame History