Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/jinseo-jang/antigravity-plugin-cc/implementgit clone --depth 1 https://github.com/jinseo-jang/antigravity-plugin-ccWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/commands/jinseo-jang/antigravity-plugin-cc/implement)<a href="https://agentmods.dev/commands/jinseo-jang/antigravity-plugin-cc/implement"><img src="https://agentmods.dev/badge/commands/jinseo-jang/antigravity-plugin-cc/implement.svg" alt="Measured on agentmods" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00012 | $0.00893 |
| Opus 5 | $0.00006 | $0.00447 |
| Sonnet 5 | $0.00002 | $0.00179 |
| Haiku 4.5 | $0.00001 | $0.00089 |
Grade A, and why
implement scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 4d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
What it actually says
Start an implement session. The Antigravity daemon runs the Antigravity SDK worker under supervision, enforcing policy, collecting git diffs, and returning a digest.
Continuity flags:
--background— start the run detached and return asession_idimmediately, skipping the watch loop. Retrieve results later with/agy:status <id>,/agy:events <id>, or/agy:watch <id>.--resume [session_id]— continue a prior run so the worker sees its earlier trajectory. Bare--resumeresumes the latest run; put the task text first and--resumelast to keep the task intact.--fresh— force a brand-new conversation (overrides--resume).
--model, --effort, and --file work on all variants.
When --file is an image, the worker should look at the image directly rather than writing throwaway shell scripts to inspect it — unnecessary scripts trigger extra approval prompts.
!python "${CLAUDE_PLUGIN_ROOT}/scripts/cao-companion.py" --plugin-data "${CLAUDE_PLUGIN_DATA}" session.implement "$ARGUMENTS"
The command above prints a JSON object with a session_id.
Recovery on error: if the companion returns a JSON-RPC error (e.g. -32602
for an incompatible --model/--effort/location combo), do NOT just print it.
Present the error message and its Options via the AskUserQuestion tool,
then re-run session.implement with the corrected args.
Foreground (no --background): supervise the session until it finishes,
using the Antigravity companion via the Bash tool:
- Watch: run
python "${CLAUDE_PLUGIN_ROOT}/scripts/cao-companion.py" --plugin-data "${CLAUDE_PLUGIN_DATA}" session.wait <session_id>. It blocks up to ~25s and returns a pending approval, "running", or "finished". - If it reports "running", run
session.wait <session_id>again. - If it reports a pending approval (a shell command needs a decision), call the
AskUserQuestion tool showing the exact command, with four options:
Approve once, Approve for this project, Approve always, and Deny.
Then apply the user's choice by running the companion:
- Approve once:
... cao-companion.py session.approve <call_id> - Approve for this project:
... cao-companion.py session.approve <call_id> project - Approve always:
... cao-companion.py session.approve <call_id> global - Deny:
... cao-companion.py session.deny <call_id>"For this project" and "always" remember the EXACT command so identical future commands auto-approve without prompting. Then return to step 1.
- Approve once:
- When it reports the session finished, run
... cao-companion.py session.events <session_id>and show the user the digest.
Never approve or deny on your own — the user decides every approval via AskUserQuestion.
Background (--background): do NOT watch. Print the returned session_id
and the retrieval hints (/agy:status <id>, /agy:events <id>, /agy:watch <id>)
and stop. If a non-allowlisted shell command is hit while unattended, the session
suspends; /agy:status <id> prints the paste-ready /agy:approve <id> [project|global]
line. The 5-minute approval timeout auto-denies — for long unattended runs,
pre-allowlist expected commands with "Approve for this project" / "Approve always".
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 4d ago First seen · 62 lines · 12 tokens per session scan A deb69ba7cf74
implement is a command published in the GitHub repository jinseo-jang/antigravity-plugin-cc (6 stars, last pushed 21d ago), licensed Apache-2.0. It adds 12 tokens to every session and 893 once invoked, about $0.0001 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other commands, from other repositories
git
Git operations with intelligent commit messages and workflow optimization.
checklist
Generate a custom checklist for the current feature based on user requirements.
clarify
Identify underspecified areas in the current feature spec by asking up to 5 highly targeted clarification questions and encoding answers back into the spec.
specify
Create or update the feature specification from a natural language feature description.
analyze
Perform a non-destructive cross-artifact consistency and quality analysis across spec.md, plan.md, and tasks.md after task generation.
converge
Assess the current codebase against the feature's spec, plan, and tasks, then append any remaining unbuilt work as new tasks to tasks.md so implement can complete it.