Skip to content
Back to skills

Agy Cli Runtime

ASecurity

Internal helper contract for calling the agy-companion runtime from Claude Code

  • 5 stars
  • 0 votes
  • 0 copies
  • 1 view
  • Added October 3, 2026
ai-agentsgobashnode

Works with

  • claude code
  • cli

Security analysis

A100/100

Scanned October 3, 2026

npx -y skills add Vit129/agy-plugin-cc --skill agy-cli-runtime --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Agy Cli Runtime?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Agy Cli Runtime
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/vit129-agy-cli-runtime/badge)](https://www.skillsdirectory.com/skills/vit129-agy-cli-runtime)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: agy-cli-runtime
description: Internal helper contract for calling the agy-companion runtime from Claude Code
user-invocable: false
---

# agy Runtime

Use this skill only inside the `agy:agy-rescue` subagent.

Primary helper:
- `node "${CLAUDE_PLUGIN_ROOT}/scripts/agy-companion.mjs" task "<raw arguments>"`

Execution rules:
- The rescue subagent is a forwarder, not an orchestrator. Its only job is to invoke `task` once and return that stdout unchanged.
- Do not call `setup`, `review`, `adversarial-review`, `status`, `result`, or `cancel` from `agy:agy-rescue`.
- Use `task` for every rescue request, including diagnosis, planning, research, and explicit fix requests.
- You may use the `gemini-3-prompting` skill to rewrite the user's request into a tighter agy prompt before the single `task` call.
- That prompt drafting is the only Claude-side work allowed. Do not inspect the repo, solve the task yourself, or add independent analysis outside the forwarded prompt text.

Command selection:
- Use exactly one `task` invocation per rescue handoff.
- If the forwarded request includes `--background` or `--wait`, treat that as Claude-side execution control only. Strip it before calling `task`, and do not treat it as part of the natural-language task text.
- If the forwarded request includes `--resume`, strip that token and add `--resume-last` to the `task` call.
- If the forwarded request includes `--fresh`, strip that token and do not add `--resume-last`.
- If the forwarded request includes `--sandbox`, pass it through to `task`.
- `--resume`: always use `task --resume-last`, even if the request text is ambiguous.
- `--fresh`: always use a fresh `task` run, even if the request sounds like a follow-up.
- If the request sounds like a follow-up ("continue", "keep going", "resume", "apply the top fix", "dig deeper"), add `--resume-last` unless `--fresh` is present.

Safety rules:
- Default to write-capable agy work in `agy:agy-rescue` unless the user explicitly asks for `--sandbox`.
- Preserve the user's task text as-is apart from stripping routing flags.
- Do not inspect the repository, read files, grep, monitor progress, poll status, fetch results, cancel jobs, summarize output, or do any follow-up work of your own.
- Return the stdout of the `task` command exactly as-is.
- If the Bash call fails or agy cannot be invoked, return nothing.

## Extended agy CLI Passthrough

The companion also supports local agy CLI features verified against `agy 1.0.16`:

- `models [--json]` lists available agy models.
- `agents [--json]` lists available agy agents (raw `agy agents` passthrough).
- `doctor [--json]` verifies plugin manifest, host wiring, agy binary/auth, and model listing.
- `changelog [--json]` prints agy release notes (no agy task run, just `agy changelog`).
- `task` accepts `--model <name>`, `--effort <low|medium|high>`, `--conversation <id>`, `--project <id>` or `--new-project` (mutually exclusive), `--dangerously-skip-permissions`, repeatable `--add-dir <path>`, `--log-file <path>`, and `--print-timeout <duration>`.
- `review` and `adversarial-review` accept the same `--model`, `--effort`, `--project`/`--new-project`, `--dangerously-skip-permissions`, `--add-dir`, `--log-file`, and `--print-timeout` flags.
- `--model` is validated against a live `agy models` call before the run starts — an unrecognized name fails fast with the current model list instead of reaching agy.
- `--effort` is validated against the static `low|medium|high` enum agy itself defines — no CLI call needed for this one.
- `task`/`review`/`adversarial-review` accept `--dry-run`: prints the resolved options, the live `agy models` list (selected model marked), and current Antigravity quota (5-hour + weekly, Gemini and Claude/GPT pools) — then exits without calling agy. Use it to pick a model and confirm there's enough quota before committing to a real run.
- Quota comes from the separately-installed `codexbar` CLI (`codexbar usage --provider antigravity --json`), not from `agy` itself — `agy`'s own `/usage`/`/quota` are TUI-only and error under `--print`/non-interactive mode. If `codexbar` isn't installed, `--dry-run` shows an install hint instead of quota and everything else still works.

Treat these as routing flags. Strip them from natural-language prompt text before invoking the helper.

`--dangerously-skip-permissions` disables all agy tool-permission prompts for that run — only forward it when the user explicitly asked for it, never by default.

`/goal <...>` is not a CLI flag — it is a prompt-level directive agy's own agent parses out of the `--print` text. Passing a prompt that starts with `/goal` through `task` already works with no companion changes; it runs agy in autonomous goal mode (uncapped duration since agy 1.0.14) rather than a single turn.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.