Agent skill

Gr2 Gl Loop

by techomancer in techomancer/iris

Autonomous graphics loop on IRIX 6.5.22 for GR2 (XZ/Extreme) and IMPACT (Solid/High/Maximum, MGRAS) - boot IRIS, build and run test/gltest (glprim, gltest) in the guest over telnet, capture board…

BSD-3-ClauseAuto-check passed

Install Gr2 Gl Loop

skills CLI
$ npx skills add techomancer/iris --skill gr2-gl-loop -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install techomancer/iris gr2-gl-loop --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/techomancer/iris.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/gr2-gl-loop .claude/skills/gr2-gl-loop && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
gr2-gl-loop
GitHub stars
125
Token cost
~3.4k tokens
SKILL.md length
1,717 words
Files
1
Skills in repo
1
Repo updated
First seen
Licence
BSD-3-Clause

At a glance

Autonomous graphics loop on IRIX 6.5.22 for GR2 (XZ/Extreme) and IMPACT (Solid/High/Maximum, MGRAS) - boot IRIS, build and run test/gltest (glprim, gltest) in the guest over telnet, capture board…

  • Works in 8 steps: cargo build --release --features… → Start it with the Bash tool's… → Start the CPU through the MCP:… → …
  • Any autonomous gfx work: GL pipeline configurations (primitives
  • SKILL.md covers Pieces, IMPACT (MGRAS) on the Indigo2…, Boot and Build glprim in the guest, plus 8 more sections
  • Calls sh and cargo; needs IRIS_ROOT_PASSWORD

What it does

Gr2 Gl Loop is an agent skill from techomancer/iris. Autonomous graphics loop on IRIX 6.5.22 for GR2 (XZ/Extreme) and IMPACT (Solid/High/Maximum, MGRAS) - boot IRIS, build and run test/gltest (glprim, gltest) in the guest over telnet, capture board traces around each run, verify rendered pixels, record/replay sessions. Use for any autonomous gfx work: GL pipeline configurations (primitives, shading, vertex/colour forms, depth, scissor, viewport), board bring-up, visuals, performance.

Its SKILL.md is about 3.4k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It works with Model Context Protocol. The licence is BSD-3-Clause.

When your agent uses it

  • Any autonomous gfx work: GL pipeline configurations (primitives
  • Vertex/colour forms

Example prompts

  • “/gr2-gl-loop”

Workflow steps

8 steps, taken from the first numbered list in SKILL.md.

  1. cargo build --release --features developer (developer: monitor tooling;
  2. Start it with the Bash tool's run_in_background: true, as a command of
  3. Start the CPU through the MCP: run_command("cpu start") (check with
  4. tools/irix-sh.py wait --timeout 900 returns once telnet answers with a
  5. tools/irix-sh.py mount-shared.
  6. GL needs a logged-in desktop (xdm's clogin holds the display otherwise).
  7. The screen saver (haven + /usr/sbin/bongo, octahedra) takes over the
  8. Stopping iris: kill it by its exact pid (`ps -eo pid,args | grep

What it can do on your machine

Read from SKILL.md and the folder at commit 0622008. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • sh
    • cargo

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names these keys or tokens, usually read from environment variables:

    • IRIS_ROOT_PASSWORD

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Gr2 Gl Loop loads about 3.4k tokens when it runs. Until then it costs about 112 tokens; SKILL.md has 1,717 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~112
When it runs · the whole SKILL.md, loaded when a task matches
~3.4k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from techomancer/iris at commit 0622008, republished under its BSD-3-Clause licence (© techomancer). 1,717 words, ~3,423 tokens.

Download SKILL.mdSave it as .claude/skills/gr2-gl-loop/SKILL.md (or your agent's skills folder).
name
gr2-gl-loop
description
Autonomous graphics loop on IRIX 6.5.22 for GR2 (XZ/Extreme) and IMPACT (Solid/High/Maximum, MGRAS) - boot IRIS, build and run test/gltest (glprim, gltest) in the guest over telnet, capture board traces around each run, verify rendered pixels, record/replay sessions. Use for any autonomous gfx work: GL pipeline configurations (primitives, shading, vertex/colour forms, depth, scissor, viewport), board bring-up, visuals, performance.

GR2 / IMPACT OpenGL test loop

Written for GR2 first; everything carries over to IMPACT with the board's own monitor commands (mgras ... for gr2 ...). See "IMPACT (MGRAS)" below for what differs.

Drive one GL configuration at a time through the real IRIX GL stack (libglcore EXPRESS -> HQ2 FIFO -> our HLE in src/dev/gr2/{hq2,gl}.rs) and check the pixels. Each glprim run draws exactly one primitive, so its FIFO trace is short and every option flips one piece of state or one address bit.

Pieces

WhatWhere
Test apptest/gltest/glprim.c (--help lists every option), built by test/gltest/irix.sh with MIPSpro cc
Guest shelltools/irix-sh.py (telnet via host port 2323, root, no password; single-shot)
File exchangebuilt-in NFS: host ./shared = guest /shared (mount with tools/irix-sh.py mount-shared). noexec: copy to /tmp before building/running
Monitoriris MCP (mcp__iris__run_command), i.e. the monitor on 127.0.0.1:8888
Token mapignore/gr2/HQ2.h "OpenGL token addressing" (source of truth; add findings there)

IMPACT (MGRAS) on the Indigo2 R10000 (IP28)

  • Config ignore/ip28/iris.toml ([graphics] board = "solidimpact" | "highimpact" | "maximpact"), disk ip28_6522_realhw.raw written directly (never kill iris while IRIX runs). Root has a password: p00pp00p. tools/irix-sh.py takes --password or IRIS_ROOT_PASSWORD=p00pp00p (glprim-case.py passes the environment on).
  • The user builds target/release/iris with --features gr4-jit,jitv2,lightning,opcodefusion,rex-jit: do not overwrite it. Build your own with --target-dir target/claude (same features) and run target/claude/release/iris --config ignore/ip28/iris.toml from the repo root (the Bash cd wrapper is broken: cd ... && exec ... inside one command works, as does run_in_background). The CPU starts by itself; this build exits on guest power-off, so after shutdown just wait for the pid to go.
  • Monitor from the shell: tools/mon.py "mgras status" "rss regs" ... (the MCP run_command has been seen returning the previous command's output). Useful: mgras status|xmap|pix X Y|stats|jit|fbdump DIR|shot F.png|trace F [hq,rss,cpu,tex|all]|rec F|rec off, rss regs (decodes the IR, shows each raster engine's backlog), locks (who holds what).
  • /tmp is emptied at boot: re-copy and rebuild gltest each boot (cp /shared/gltest/* /tmp/gltest/ && sh irix.sh gltest glprim). /shared mounts by itself on this image.
  • Desktop readiness: telnet comes up first; wait for 4Dwm and then toolchest in ps -ef, plus ~15 s, before screenshots; then xset s off.
  • tools/glprim-case.py detects the board. IMPACT's scanout applies the DAC gamma (SGI default 1.7), so mid-level colours read back brighter than GR2's (the pink clear 1, 0.5, 0.75 shows as ffaad7); probe saturated colours, or check values with glprim --read ... --check. glprim picks the deepest visual by default (12,12,12, which the raster JIT leaves to the interpreter): use --visual 8,8,8,8 for the JIT path.
  • Textures: tools/gltex-matrix.py runs glprim's tex scene over formats and sizes and checks every texel's raw framebuffer word (all pass as of 48089c43). Several GL clients share TRAM through the texture manager's page saves and restores (tl_spec 0xd60 loads in mgras trace F hq,tex): a bug there damages every client.
  • Performance: board-bound runs slow the guest clock (lost ticks), so guest-side gltest --bench is meaningless (see rules/mgras/guest-gl-benchmarks-miss-board-time.md). Record the session (mgras rec F ... mgras rec off) and time its replay offline with MGRAS_REPLAY_NOCHECK=1 MGRAS_REPLAY_BOARD=high|max MGRAS_REPLAY=F cargo test --release --target-dir target/claude --lib --features gr4-jit mgras_replay_file -- --nocapture (wall time + final state hash).
  • Goldens: MGRAS_REPLAY=ignore/mgras-golden/desktop2.rec must pass every checkpoint (also with MGRAS_REPLAY_BOARD=max). irix6522-boot.rec is stale (11 of 16 checkpoints fail at 35a40840 too).
  • Hardware findings go in ignore/gr4/*.h (its own git repo; the user commits it), design notes in rules/mgras/.

Boot

  1. cargo build --release --features developer (developer: monitor tooling; do not combine with lightning).
  2. Start it with the Bash tool's run_in_background: true, as a command of its own (target/release/iris > <scratchpad>/iris-run.log 2>&1). A nohup ... & inside a chained command does not survive the command, and starting it right after killing the old process can fail: check pgrep -a iris before going on. (config: iris.toml, Indy R5000 + XZ, disk scsi1.raw.)
  3. Start the CPU through the MCP: run_command("cpu start") (check with get_status if unsure).
  4. tools/irix-sh.py wait --timeout 900 returns once telnet answers with a login prompt (network and inetd are up; X/xdm come up around then too).
  5. tools/irix-sh.py mount-shared.
  6. GL needs a logged-in desktop (xdm's clogin holds the display otherwise). root autologin is set up: /etc/autologin = root, and the ONE-SHOT flag /etc/autologin.on, which clogin deletes when it logs in. Re-create it (touch /etc/autologin.on) every boot before shutdown. The session takes ~1 min after telnet comes up: wait until ps -ef | grep 4Dwm shows it.
  7. The screen saver (haven + /usr/sbin/bongo, octahedra) takes over the whole screen after 10 min idle and every probe reads black. The desktop re-applies its saver settings after /root/.sgisession, so run /usr/bin/X11/xset s off once the session is up, and check xset q | grep timeout: = 0.
  8. Stopping iris: kill it by its exact pid (ps -eo pid,args | grep '[t]arget/release/iris$'). pgrep -f target/release/iris also matches the shell running the command and kills it.

iris.toml currently writes the disk image directly (no overlay): never kill iris while IRIX is running (see Shutdown).

Build glprim in the guest

mkdir -p shared/gltest && cp test/gltest/{glprim.c,main.c,irix.sh} shared/gltest/
tools/irix-sh.py run 'mkdir -p /tmp/gltest && cp /shared/gltest/* /tmp/gltest/' \
                     'cd /tmp/gltest && sh irix.sh glprim 2>&1'

Every session gets DISPLAY=:0.0, /bin/sh, cwd /tmp. The exit status of irix-sh.py run is the last command's status (124 = timeout/connection).

One test case

Preferred: tools/glprim-case.py CASE [--probe gx,gy[=rrggbb]]... [--notrace] -- <glprim options> does all of the below (trace, run, window origin, probes in GL window coordinates, fbdump, crop of the window to $GLPRIM_OUT/CASE.png, trace off, kill glprim) and exits with the number of failed probes. It talks to the monitor on 127.0.0.1:8888 directly (works alongside the MCP). Stage the sources first (cp test/gltest/* shared/gltest/, then copy to /tmp in the guest and sh irix.sh glprim).

By hand:

run_command("gr2 trace /tmp/gl-<case>.log all")        # MCP
tools/irix-sh.py run 'cd /tmp/gltest && ./glprim <options> --hold 4000'
run_command("gr2 fbdump /tmp/gl-<case>")                # during the hold, or
run_command("gr2 pix <x> <y>")                          # spot checks (display coords)
run_command("gr2 trace off")

Run the capture commands while --hold keeps the window up (launch the guest command in the background with & if you need to sample during the hold). glprim prints its configuration and its real window origin (window origin OX,OY size WxH; the window manager moves windows). A GL window pixel (gx, gy) is display pixel (OX + gx, OY + H - 1 - gy).

Default geometry (unit square scaled to the viewport, glOrtho 0..w, 0..h): triangle (0.2w,0.2h) red, (0.8w,0.2h) green, (0.5w,0.8h) blue. Flat shading uses the provoking vertex (last vertex for triangles/strips/fans/quads, the FIRST vertex for GL_POLYGON). Clear default is pink (1, 0.5, 0.75) = rgb=ff80bf.

Show full SKILL.md (746 more words)Show less

Verify

  • gr2 pix x y prints the VRAM word, DID, XMAP mode and composed rgb.
  • gr2 fbdump DIR writes VRAM planes and the composed screen as PNGs; view them with the Read tool.
  • Expected values: compute from the GL spec (pixel centres at +0.5, fill rule top-left style; flat = provoking vertex colour; smooth = barycentric interpolation, allow +-2 per channel). Check interior, each edge's first in/out pixel, and a pixel just outside the window (must be untouched).
  • In the trace, HQ exec lines show the HLE's view: VERTEX (...) -> window (...), GL triangle ..., GL_CLEAR ... rect. Unknown tokens appear as tokNNN / (N words, not implemented) with their modifier bits decoded (V3|USEV:VERTEX, C3:COLOR, LOADV:...).

Iterate

--scene depth|stencil|alphatest|blend draws small multi-primitive scenes with known images (see the comment above draw_scene in glprim.c); verified probes are in the history of this skill's cases (depth: (170,150) red, (230,150) blue; stencil: (200,120) blue, (20,20) pink; blend: (200,120) 800080, (300,120) 8040df; alphatest: (200,200) ~1c1cc6, (90,62) pink).

Change one option at a time. Suggested order: clear only (-p none), flat triangle, --smooth, --mono, every -p primitive, --vtx 2f/4f/2i/3i/2s/3d, --color 4f/3ub/4ub/once, --viewport, --scissor, --cull/--cw, --polymode, --persp, --depth, --db. When a new token or format shows up, decode it against ignore/gr2/libglcore/decomp/EXPRESS (struct offset X = FIFO index (X - 0x2000)/4), record it in HQ2.h with the emitter as citation, implement in src/dev/gr2/gl.rs, add a replay test in src/dev/gr2/hq2_tests.rs built from the trace (see glprim_triangle), and re-run the case.

Offline replay of real streams

A guest restart costs ~6 minutes, so capture once and iterate offline: extract the HQ2 FIFO writes of a case (from the window setup 0x1e5 to the GL_FINISH after the draw, 2D command groups 0x12C..0x1EA dropped) into src/dev/gr2/testdata/glprim_<scene>.trace ("index value" hex per line) and replay it in a unit test with replay_trace() (hq2_tests.rs); the window origin is the first two words after 1e5. See the lighting/fog tests.

Demos

/usr/demos/General_Demos/<name>/<name> (atlantis, ideas, buttonfly, flight, powerflip, ...). GL state is swapped per context on GE_HQMSAV; still, test demos one at a time.

Demos that wait for input need it injected: ideas sits on its title screen until clicked (and needs another click to restart after its animation), so a run without a click only ever draws the title.

Driving input (mouse, keyboard)

Two ways, both scriptable:

  • Monitor, ps2 (any PS/2 machine: Indy, Indigo2/IP28): ps2 mouse <dx> <dy> [buttons] sends one relative mouse packet (buttons: 1 left, 2 right, 4 middle; send a packet with the button, then one with 0 to release), ps2 type <ascii text> types a string, ps2 enter presses Enter, ps2 status shows the controller state. Moves are relative and go through X's pointer acceleration: to reach a spot, first push the pointer into the top-left corner with a few large negative moves, then step to the target in small moves.
    run_command("ps2 mouse -500 -500")  # corner (repeat to be sure)
    run_command("ps2 mouse 40 30")      # small steps to the target
    run_command("ps2 mouse 0 0 1")      # press left
    run_command("ps2 mouse 0 0 0")      # release
  • In the guest, XTest: test/gltest/xclick.c (xclick NAME [BUTTON [X Y]] or xclick NAME key KEYSYM) clicks or presses a key at exact window coordinates in the first window whose name starts with NAME; test/gltest/xresize.c (xresize NAME W1 H1 W2 H2 [COUNT] [DELAY_MS]) resizes a window repeatedly. Build with sh irix.sh xclick xresize. Precise, no acceleration, but needs a telnet session.

Demo sources (ideas, atlantis, ...) are published on opengl.org: read them before guessing what a demo sends.

Shutdown

tools/irix-sh.py run 'touch /etc/autologin.on; sync; shutdown -p -g0 -y'

Then wait about a minute (the developer build does not exit on power-off), check /tmp/iris-run.log / the monitor for the power-off, and only then kill the iris process.

Gotchas

  • The host shell's cd may be aliased; use builtin cd or absolute paths.
  • /shared is noexec: always build and run from /tmp.
  • Stock IRIX has no bash (scripts must be /bin/sh), no pkill, and chmod wants a+x rather than +x.
  • GL state is per context: GE_HQMSAV (0x1F0, kernel Gr2PcxSwap) swaps the HLE's GlState by context id; state 0 = new context = defaults. A trace that starts mid-session has no switch for the running client: replays keep the live state for it.
  • MIPSpro is C89-ish: declarations at block start, no <stdint.h>.
  • A GL client that waits forever on version bit 0 (FIN3) is waiting for a Finish (0xA3) we did not answer; a stuck client on bit 1 is FIN2 (kernel context switch).
  • Window clipping arrives as 0x1E5 only on the switch into a GL context after its clip changed (see rules/gr2/gl-window-clipping.md). To trace clipping, start the trace before moving the covering window.

© techomancer, BSD-3-Clause. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in .claude/skills/gr2-gl-loop of techomancer/iris.

Open the folder on GitHubat commit 0622008

Compare with similar skills

Gr2 Gl Loop next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Gr2 Gl Loop compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Gr2 Gl Loop this skilltechomancer/iris125—~3.4kAutomated safety check: PassBSD-3-Clause
MCP Server Builderanthropics/skills180k63 repos~2.3kAutomated safety check: PassApache-2.0
MCP Server BuildershareAI-lab/learn-claude-code78k4 repos~1.2kAutomated safety check: PassMIT
MCP Integration for Pluginsanthropics/claude-plugins-official38k11 repos~3.1kAutomated safety check: PassApache-2.0
Figma use_figma Plugin API Ruleswarpdotdev/warp65k4 repos~4.4kAutomated safety check: PassAGPL-3.0
Stitch to Remotion Walkthrough Videosgoogle-labs-code/stitch-skills8.5k6 repos~3.2kAutomated safety check: NotesApache-2.0

Similar skills

  • MCP Server Builder

    anthropics/skills

    Official

    Guides the design and implementation of Model Context Protocol servers in TypeScript or Python, from tool naming and error messages to evaluation.

    180k GitHub starsUsed in 63 repos~2.3k tokens
    Agent WorkflowsAuto-check passed
  • MCP Server Builder

    shareAI-lab/learn-claude-code

    Walks through building MCP servers in Python or TypeScript that expose tools, resources and prompts to Claude, with templates, registration and testing.

    78k GitHub starsUsed in 4 repos~1.2k tokens
    Agent WorkflowsAuto-check passed
  • MCP Integration for Plugins

    anthropics/claude-plugins-official

    Official

    Explains how to bundle Model Context Protocol servers in a Claude Code plugin, covering config files, stdio, SSE, HTTP and WebSocket server types, and authentication.

    38k GitHub starsUsed in 11 repos~3.1k tokens
    Agent WorkflowsAuto-check passed
  • Required groundwork before any use_figma call: the rules and reference files for running JavaScript in a Figma file through the Plugin API without common failures.

    65k GitHub starsUsed in 4 repos~4.4k tokens
    Frontend & DesignAuto-check passed
  • Stitch to Remotion Walkthrough Videos

    google-labs-code/stitch-skills

    Official

    Builds walkthrough videos from Stitch design projects using Remotion, with transitions, zoom effects and text overlays on each screen.

    8.5k GitHub starsUsed in 6 repos~3.2k tokens
    Media & CreativeAuto-check: notes
  • MCP Development

    coollabsio/coolify

    A skill your agent uses for Laravel MCP development. An agent skill from coollabsio/coolify.

    63k GitHub starsUsed in 1 repo~949 tokens
    Frontend & DesignAuto-check passed

Questions about Gr2 Gl Loop

What does Gr2 Gl Loop do?

Autonomous graphics loop on IRIX 6.5.22 for GR2 (XZ/Extreme) and IMPACT (Solid/High/Maximum, MGRAS) - boot IRIS, build and run test/gltest (glprim, gltest) in the guest over telnet, capture board…. Gr2 Gl Loop is an agent skill from techomancer/iris.22 for GR2 (XZ/Extreme) and IMPACT (Solid/High/Maximum, MGRAS) - boot IRIS, build and run test/gltest (glprim, gltest) in the guest over telnet, capture board traces around each run, verify rendered pixels, record/replay sessions.

When should I use Gr2 Gl Loop?

Gr2 Gl Loop fits situations like: any autonomous gfx work: GL pipeline configurations (primitives; vertex/colour forms.

How do I install Gr2 Gl Loop in Claude Code?

Run `npx skills add techomancer/iris --skill gr2-gl-loop -a claude-code`. Or copy the skill folder (.claude/skills/gr2-gl-loop in techomancer/iris) into .claude/skills/gr2-gl-loop in your project. Claude Code loads it when a task matches its description.

How do I install Gr2 Gl Loop in Codex?

Run `npx skills add techomancer/iris --skill gr2-gl-loop -a codex`. Or copy the skill folder (.claude/skills/gr2-gl-loop in techomancer/iris) into .agents/skills/gr2-gl-loop in your project. Codex loads it when a task matches its description.

Can I use Gr2 Gl Loop in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add techomancer/iris --skill gr2-gl-loop -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/gr2-gl-loop, .gemini/skills/gr2-gl-loop, .github/skills/gr2-gl-loop and .opencode/skills/gr2-gl-loop in your project.

What does Gr2 Gl Loop need to run?

Going by SKILL.md and its folder, Gr2 Gl Loop needs the command-line tools its instructions call (sh and cargo) and credentials named IRIS_ROOT_PASSWORD.

Does Gr2 Gl Loop access the network?

SKILL.md contains no URLs. Any network use would come from the scripts or tools the agent runs. This is read from the text; nothing was executed.

Is Gr2 Gl Loop safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Gr2 Gl Loop use?

Gr2 Gl Loop is published under the BSD-3-Clause licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Gr2 Gl Loop use?

About 3.4k tokens (SKILL.md is roughly 14k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Gr2 Gl Loop?

Skills that share tags, products or a category with Gr2 Gl Loop: MCP Server Builder (anthropics/skills, 180k stars), MCP Server Builder (shareAI-lab/learn-claude-code, 78k stars), MCP Integration for Plugins (anthropics/claude-plugins-official, 38k stars) and Figma use_figma Plugin API Rules (warpdotdev/warp, 65k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Gr2 Gl Loop?

techomancer (a GitHub user) maintains it in techomancer/iris, which has 125 GitHub stars. The repository was last updated on October 11, 2026.

Source: techomancer/iris on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.