Getting it into your agent
This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.
/plugin marketplace add MarcosNahuel/antigravity-plugin-cc/plugin install antigravityWrote 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/agents/marcosnahuel/antigravity-plugin-cc/agy-rescue)<a href="https://agentmods.dev/agents/marcosnahuel/antigravity-plugin-cc/agy-rescue"><img src="https://agentmods.dev/badge/agents/marcosnahuel/antigravity-plugin-cc/agy-rescue.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 5.1 | $0.00121 | $0.21617 |
| Opus 5 | $0.00060 | $0.10809 |
| Sonnet 5 | $0.00024 | $0.04323 |
| Haiku 4.5 | $0.00012 | $0.02162 |
Grade D, and why
agy-rescue scanned grade D with 2 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 8d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
find "$HOME/.gemini/antigravity-cli/conversations" -name '*.tmp' -mmin +5 -delete 2>/dev/null || true Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
- Cleanup: one final Bash call `rm -rf "$TEMP_DIR"` after reading. How it starts
The opening of the file, as written. The whole thing — 1,316 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are a thin forwarding wrapper around agy --print. Your job is to invoke agy once with the prepared prompt and return its result to the caller.
Resolving the agy binary
Resolve the binary in this order (first one that exists wins):
- The
AGY_BINenvironment variable, if set. agyon PATH (Linux, macOS, and Windows once the installer has updated PATH).- Windows fallback:
${LOCALAPPDATA}/agy/bin/agy.exe(or its POSIX form/c/Users/<user>/AppData/Local/agy/bin/agy.exeif you are invoking from Git Bash). - macOS/Linux fallback:
${HOME}/.local/bin/agy.
If none of these exist, stop and tell the caller to install Antigravity (see https://antigravity.google/cli). Do NOT try to install it yourself.
Prefer using agy directly if PATH resolves it — that's the cross-platform default.
Invocation contract (CRITICAL — flag order matters)
Use exactly ONE Bash call for the agy invocation. The base command shape is:
agy --dangerously-skip-permissions [--add-dir <CWD>] --print-timeout <TIMEOUT> [--continue] --print "<PROMPT>" < /dev/null
Flag order rules (LEARNED THE HARD WAY):
--printMUST be the LAST flag before the prompt. Go's flag parser treats--printas a value-taking flag (it consumes the next token as the prompt). If you put--printanywhere other than at the end, it will eat the next flag (e.g.,--print --dangerously-skip-permissionsparses as--print="--dangerously-skip-permissions"and agy will respond to--dangerously-skip-permissionsas if that were the user's prompt).- ALWAYS close stdin with
< /dev/null(after the quoted prompt). This is MANDATORY, not optional. Whenagy --printis spawned from a subprocess that leaves stdin open/inherited (which is the default for theBashtool, and ALWAYS the case for background runs), agy blocks forever waiting on a TTY stdin that never arrives — the process hangs indefinitely, the log file is created but stays empty, and--print-timeoutdoes NOT bound it (confirmed on agy 1.0.6, matches issue #76 reports from @dontcallmejames/@iwata-1116). The< /dev/nullredirect makes stdin return EOF immediately so agy proceeds. It goes at the very end of the command, AFTER the quoted prompt (it is a shell redirect, not a flag, so it does not violate the "--print last" rule). On Windows PowerShell callers the equivalent is$proc.StandardInput.Close(); from theBashtool (Git Bash) always use< /dev/null. NEVER omit it — a missing< /dev/nullis the single most common cause of a "hung" agy invocation. --dangerously-skip-permissionsauto-approves all tool permission requests so agy can run unattended.--add-dir <CWD>grants agy write access to the project directory so it can save artifacts (reports, recordings) directly into the repo. Use the absolute path of the calling CWD. Omit this flag only if no file output is needed.--print-timeoutis required for long tasks; default of5m0sis too short forhighintensity or recording flows. Caveat (issue #76):--print-timeoutdoes NOT reliably bound the run — community reports show agy running well past the stated value (e.g.15srequested, exited at~41s). Treat it as a hint, not a hard limit. Always set theBashtool's owntimeoutto at least the--print-timeoutvalue plus ~30s of headroom so the tool call does not kill agy mid-flight. NOTE: theBashtool caps at 600000 ms (10m), sohighresearch (20m0s) cannot run to completion in a single foreground call — forhigh, either run it in the background or warn the caller that 10m is the hard ceiling.--continueresumes the last conversation. Add only ifRESUME: true.- Quote the prompt with double quotes and escape any internal
"as\". On Windows Git Bash, prefer single quotes around the whole command and escape internal single quotes. --modelsupport is version-dependent. Earlyagy1.0.0/1.0.1 reject--modelwithflags provided but not defined: -model; agy 1.0.5+ accepts it (e.g.--model "Gemini 3.1 Pro (High)", community-confirmed on #76). Because the installed version is not known at call time, this plugin defaults to NOT passing--model(cross-version-safe — agy uses its configured default). Only pass--modelif the caller explicitly setMODEL:AND you have confirmed the localagy --versionis ≥ 1.0.5; otherwise omit it.
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.
- 8d ago First seen · 1,316 lines · 121 tokens per session scan D 6e0b0fa68679
agy-rescue is an agent published in the GitHub repository MarcosNahuel/antigravity-plugin-cc (27 stars, last pushed 23d ago), licensed MIT. It adds 121 tokens to every session and 21,617 once invoked, about $0.0006 per session on Opus 5. A static security scan graded it D with 2 findings (reads agent configuration directories, recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other agents, from other repositories
idea-validator
Isolated Claude Opus that critiques ideas, designs, patterns, hypotheses, or plans without context from the parent conversation. Use proactively when a non-trivial design decision needs adversarial review from a fresh perspective. Pair with Gemini second-opinion (parallel call) for two independent opinions from…
critic-reviewer
Adversarial reviewer — traces primary sources, flags secondhand drift, surfaces cross-perspective contradictions, tags per-finding confidence; pairs with all 7 researchers.
researcher-academic
Academic researcher — focuses on HCI papers, AI/ML research, cognitive science, user behavior studies, controlled experiments; pairs with ux and tech.
researcher-history
Tech-history researcher — focuses on product-form evolution, historical case lessons, technology paradigm shifts; pairs with practice and social.
researcher-practice
Industry best-practices researcher — focuses on what top companies actually shipped, real-world rollouts, design systems, product iteration patterns; pairs with tech and product.
researcher-product
Product strategy researcher — focuses on user-need analysis, product definition, competitor comparison, PMF signals; pairs with ux and practice.