Skillget

agent-supagents

agent
Guvenlik Denetimi
Uyari
Health Uyari
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 6 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

Compile a single AI supagent into multiple AI subagents (Claude Code, Gemini CLI, GitHub Copilot, Cursor, OpenCode, Kilo Code).

README.md

Supagents

Define a role once, then compile it into native subagent profiles for nine coding harnesses. Keep instructions shared and model, tool, and permission settings explicit for each host.

uv tool install supagents
supagents init reviewer --project
# Edit .agents/supagents/reviewer.md: write the instructions and keep the desired target blocks.
supagents build --project
supagents build --project --check

Requires Python 3.12+. See the changelog for release changes.

Target Harness Project output
AGY Antigravity .agents/agents/<name>.md
CLAUDE Claude Code .claude/agents/<name>.md
CODEX Codex .codex/agents/<name>.toml
COPILOT GitHub Copilot .github/agents/<name>.agent.md
CURSOR Cursor .cursor/agents/<name>.md
GEMINI Gemini CLI .gemini/agents/<name>.md
GROK Grok Build .grok/agents/<name>.md
KILO Kilo Code .kilo/agents/<name>.md
OPENCODE OpenCode .opencode/agents/<name>.md

One source, native settings

---
name: reviewer
description: Review assigned changes and report demonstrated defects.
CLAUDE:
  model: inherit
  tools: Read, Glob, Grep
CODEX:
  sandbox_mode: read-only
AGY:
  subagent: true
  model: inherit
  inheritMcp: false
  tools: [view_file, list_dir, find_by_name, grep_search]
---

Read the assigned files and requirements. Return findings with file and line references, evidence, and unresolved gaps. Do not implement fixes or delegate further work.

Only declared target blocks produce outputs. Supagents preserves the Markdown body and merges shared metadata with each target's overrides; Codex receives the body as TOML developer_instructions. It does not translate tool names, permissions, or model IDs. Native settings and parent-session policies determine runtime behavior.

Commands

Command Purpose
init NAME Scaffold a source with all bundled targets
build Validate the complete plan, then write changed outputs
build --check Fail if outputs differ; write nothing
build --dry-run Preview planned changes
list Show source-to-output mappings
clean --dry-run Preview generated files no longer produced
clean Remove those orphaned files

Use --project or --global, repeatable --target, --config, and --source-dir to select scope and locations. By default, a local .agents/supagents/ selects project scope; otherwise the CLI uses ~/.agents/supagents/.

Builds refuse to overwrite handwritten files, symlinks, sources, or colliding outputs. Cleanup refuses invalid or ambiguous sources. See usage and migration for configuration precedence, safety limits, host references, and chezmoi integration.

Development

mise trust
mise run install
mise run all

Python 3.12+ is supported; mise pins the development toolchain and uv locks dependencies. CI qualifies Linux and macOS, with an additional Python 3.12 test run. The gate audits dependencies and smoke-tests the built wheel in an isolated environment. Lefthook calls the same check and test tasks. Local checks do not establish hosted CI or live harness delegation results.

The repository also supplies an Agent Skill, Claude Code, Gemini CLI, and Copilot plugin manifests, and a downstream pre-commit hook. See plugin installation.

MIT license.

Yorumlar (0)

Sonuc bulunamadi