thirdkeyai/symbi-claude-code/symbi-policy

Claude Code and agent skill by thirdkeyai/symbi-claude-code

Trust tier 3
  • community
Source check pending

Checking that the SKILL.md file is still in the repository.

Draft and validate Cedar policies using the request schema of the selected execution route.

Install

npx vskill@latest install thirdkeyai/symbi-claude-code/symbi-policy

SKILL.md

Cedar policies

First determine whether the policy governs standalone plugin hooks or managed runtime tools. They use different action namespaces and contexts. Inspect the shipping request contract before using identity attributes or approvals.

Standalone plugin contract

The CLI receives tool_name and complete tool_input, and constructs:

  • principal: Symbi::Agent::"claude-code" (fixed by this CLI, including Cursor)
  • action: Symbi::Action::"tool_call::<native tool name>"
  • resource: Symbi::Resource::"default"
  • context: { "tool_input": <original arguments> }

For example, permit direct reads and content searches while leaving other calls denied by default:

cedar
permit(principal, action == Symbi::Action::"tool_call::Read", resource);
permit(principal, action == Symbi::Action::"tool_call::Grep", resource);

Claude Code shell calls use Bash; Cursor shell calls use Shell. This fixed principal is not a verified user identity. Local exceptions do not bypass Cedar.

Validate with complete representative payloads, including denied controls:

sh
printf '%s\n' '{"tool_name":"Read","tool_input":{"file_path":"README.md"}}' |
symbi policy evaluate --stdin --json --policies /absolute/policies

Inspect the structured decision and reason. Exit zero alone does not establish successful policy loading: the current runtime can return default allow when built without Cedar or with no policies; the plugin rejects those results. Do not use an empty {} event as evidence that a real request is permitted.

Managed runtime

Start from the policies generated by the chosen runtime profile. The managed CLI example uses Action::"tool_call::claude_code" for admission and exact tool names for brokered calls. File roots, approval and budget checks are additional runtime controls; a Cedar permit does not replace them.

Draft changes for operator review outside active policy directories. Do not modify an active control file or use shell execution to circumvent a denied write. Evaluate both allowed and forbidden representative requests before applying changes.

Security

  • Security scanPassed, 95 of 100Pattern checks on the skill's files.
  • Model reviewPassed

Quality evaluation

  • Task evaluationNot runNo evaluation yet. Run one in Skill Studio.

Source

Popularity

7-day trend score
0

Not refreshed yet.

Works with

Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Windsurf
  • Codex
  • Gemini CLI
  • Cline
  • Amp
  • Roo Code
  • Goose
  • Continue
  • Kilo Code
  • Junie
  • Trae
  • OpenHands
  • Qwen Code
  • AI Agent
  • Tabnine Chat
  • Sourcegraph Cody
  • Amazon Q
  • Firebase Genkit
  • Sweep
  • Devin
  • Phind
  • Mentat
  • Claude Desktop
  • Aider
  • GPT Pilot
  • Plandex
  • bolt.new
  • v0
  • Replit Agent
  • Anse
  • TypingMind
  • BetterChatGPT
  • LibreChat
  • Warp AI
  • Pieces
  • Fig AI
Embed badge
vskill VERIFIED badge
[![vskill VERIFIED](https://verified-skill.com/api/v1/skills/thirdkeyai/symbi-claude-code/symbi-policy/badge)](https://verified-skill.com/skills/thirdkeyai/symbi-claude-code/symbi-policy)

Something wrong with this skill or its page? Tell us.