Skillget

agentsummons

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

No AI report is available for this listing yet.

SUMMARY

Go library + CLI for invoking agent harnesses in headless mode.

README.md

agentsummons

Go library + CLI for invoking agent harnesses (Antigravity CLI, Claude
Code, Codex CLI, GitHub Copilot CLI) in headless mode.

Every headless-agent experiment rediscovers the same lore: which binary,
which permission-bypass flag, that agy ignores your working directory
without --add-dir and swallows everything after -p, that Codex wants a
bare positional prompt under an exec subcommand. agentsummons owns that
knowledge once, behind one Go API and one CLI, so your scripts and
experiments don't have to.

It is the companion to
agentminutes, which
parses the session transcripts harnesses write: agentsummons convenes
the meeting, agentminutes takes the minutes.

Install

brew install agent-ecosystem/tap/agentsummons
# or
go install github.com/agent-ecosystem/agentsummons/cmd/agentsummons@latest

Also on npm and
PyPI for Node and Python
projects, with prebuilt static binaries on the
releases page.
See Installation for all
options.

Quick start

# Invoke a harness. Default is a transparent wrapper: harness
# stdout/stderr pass through and the harness exit code propagates.
agentsummons run --harness claude-code -p "Summarize this repo" \
  --allowed-tools Read,Grep,Glob --auto-approve

# --json swaps to a single result envelope on stdout, the shape
# scripts and containers consume.
agentsummons run --harness codex -p "hi" --json

# Compare installed harness versions against the validated flag
# surface before an experiment burns tokens. Free; only version
# commands run.
agentsummons doctor

As a Go library:

res, err := agentsummons.Run(ctx, agentsummons.Request{
    Harness:     agentsummons.ClaudeCode,
    Prompt:      "Summarize this repo",
    Workdir:     dir,
    AutoApprove: true,
})

Capability matrix

One request shape, translated per harness:

Harness Binary Model Session ID Resume Allowed tools JSON output Auto-approve
antigravity agy --model "<display name>" (none) --conversation <id> (none) --output-format json (envelope) --dangerously-skip-permissions
claude-code claude --model <id> --session-id <uuid> --resume <session-id> --allowedTools a,b --output-format json (envelope) --dangerously-skip-permissions
codex codex -m <id> (none) exec resume <session-id> (none) --json (JSONL events) --dangerously-bypass-approvals-and-sandbox
copilot copilot --model <id> --session-id <uuid> --resume=<session-id> --available-tools=a,b --output-format json (JSONL events) --allow-all

agentsummons info has the full manifests, including per-harness quirk
notes. Setting a field a harness can't express is a loud
*UnsupportedError, never a silent drop, and agentsummons never applies
policy by default: permission bypass is always the caller's explicit
choice.

Documentation

Full documentation is available at
agentsummons.dev:

  • Quickstart: install
    agentsummons and invoke your first harness
  • CLI: the run, build, info, and
    doctor commands, exit codes, and the JSON contract
  • Go Library: the Run and Build
    API, error semantics, and escape hatches
  • Harness Capabilities: the
    capability matrix and per-harness quirk notes
  • Multi-Turn Sessions:
    resuming conversations across invocations, with worked examples
  • Example: An Experiment Runner:
    agentsummons driving a real multi-harness benchmark
  • Containers: baking
    reproducible eval images
  • Flag Drift: the validated
    flag surface and what a drift candidate means

License

MIT.

Reviews (0)

No results found