New: 98,000+ skills indexed from GitHub repos with 100+ stars · every skill safety-checkedBrowse top skillsTop skills
Esc to closeBrowse all skills
Submit a skillBrowse skills
Submit a skillBrowse skills

Agent skill

B200 Tmem Lifecycle Planner

by mirage-project in mirage-project/mirage

A skill your agent uses when the user is using tcgen05 on Blackwell/B200 and needs to plan TMEM accumulators, block-scale factors, TMEM column allocation, the tcgen05.ld/st/cp paths, epilogue…

Apache-2.0Auto-check passed

Install B200 Tmem Lifecycle Planner

skills CLI
$ npx skills add mirage-project/mirage --skill b200-tmem-lifecycle-planner -a claude-code

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

GitHub CLI
$ gh skill install mirage-project/mirage b200-tmem-lifecycle-planner --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/mirage-project/mirage.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/b200-tmem-lifecycle-planner .claude/skills/b200-tmem-lifecycle-planner && 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
b200-tmem-lifecycle-planner
GitHub stars
2.5k
Token cost
~1.7k tokens
SKILL.md length
778 words
Files
2
Skills in repo
24
Repo updated
First seen
Licence
Apache-2.0

At a glance

A skill your agent uses when the user is using tcgen05 on Blackwell/B200 and needs to plan TMEM accumulators, block-scale factors, TMEM column allocation, the tcgen05.ld/st/cp paths, epilogue…

  • Works in 5 steps: First compute the columns and alignment… → then map the logical tiles onto… → annotate who writes, who reads, and when… → …
  • The user is using tcgen05 on Blackwell/B200 and needs to plan TMEM accumulators
  • SKILL.md covers R — Source evidence (Reading,…, I — Methodology skeleton…, A1 — Applications in the… and A2 — Trigger scenarios (Future…, plus 4 more sections
  • Instructions only: no scripts, shell commands, URLs or credentials in SKILL.md

What it does

B200 Tmem Lifecycle Planner is an agent skill from mirage-project/mirage. Use when the user is using tcgen05 on Blackwell/B200 and needs to plan TMEM accumulators, block-scale factors, TMEM column allocation, the tcgen05.ld/st/cp paths, epilogue readback, and safe deallocation. Produces a TMEM region map, column budget, and lifecycle. Not for Hopper/Ampere or code that treats TMEM as ordinary shared memory.

Its SKILL.md is about 1.7k tokens, which your agent loads only when the skill is triggered. The skill folder holds 1 other file (for example `test-prompts.json`).

The repository describes itself as: Mirage Persistent Kernel: Compiling LLMs into a MegaKernel. The licence is Apache-2.0.

When your agent uses it

  • The user is using tcgen05 on Blackwell/B200 and needs to plan TMEM accumulators
  • Block-scale factors
  • TMEM column allocation
  • The tcgen05.ld/st/cp paths

Example prompts

  • “/b200-tmem-lifecycle-planner”

Workflow steps

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

  1. First compute the columns and alignment each object class needs;
  2. then map the logical tiles onto TLane/TCol;
  3. annotate who writes, who reads, and when each completes;
  4. arrange region reuse and the final free;
  5. ensure the epilogue's register fragments are consistent with the TMEM mapping used at write time.

What it can do on your machine

Read from SKILL.md and the folder at commit f9eb70c. 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

    No scripts in the folder and no shell commands in SKILL.md.

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

  • Network

    Links to these hosts (documentation or services it may open):

    • mlc.ai

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

  • Credentials

    Names no API keys, tokens, secrets or passwords.

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

Context cost

B200 Tmem Lifecycle Planner loads about 1.7k tokens when it runs. Until then it costs about 92 tokens; SKILL.md has 778 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~92
When it runs · the whole SKILL.md, loaded when a task matches
~1.7k

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 mirage-project/mirage at commit f9eb70c, republished under its Apache-2.0 licence (© mirage-project). 778 words, ~1,687 tokens.

Download SKILL.mdSave it as .claude/skills/b200-tmem-lifecycle-planner/SKILL.md (or your agent's skills folder). This skill also uses 1 other file; get the full folder from GitHub.
name
b200-tmem-lifecycle-planner
description
Use when the user is using `tcgen05` on Blackwell/B200 and needs to plan TMEM accumulators, block-scale factors, TMEM column allocation, the `tcgen05.ld/st/cp` paths, epilogue readback, and safe deallocation. Produces a TMEM region map, column budget, and lifecycle. Not for Hopper/Ampere or code that treats TMEM as ordinary shared memory.
source_book
Modern GPU Programming For MLSys (MLC Community) + NVIDIA Blackwell Tuning/Compatibility Guides
source_chapter
S8; S7; S14
tags
b200, blackwell
related_skills
b200-layout-contract-auditor, b200-tcgen05-mma-contract-builder, b200-mbarrier-protocol-auditor, b200-flash-attention4-planner
version
0.1.0
<!-- Distilled from "Modern GPU Programming for MLSys" — https://mlc.ai/modern-gpu-programming-for-mlsys/ -->

B200 TMEM Lifecycle Planner

R — Source evidence (Reading, paraphrased)

  • [S8] TMEM is a Blackwell-only, CTA-scoped two-dimensional Lane × Col scratchpad, with 128 Lane rows and up to 512 32-bit Cols.
  • [S8] TMEM must be explicitly allocated and freed in 32-column units; ordinary shared load/store cannot access it.
  • [S8] tcgen05.ld/st/cp are all dedicated asynchronous paths, each with its own completion rules.
  • [S7] Both the accumulator and the block-scale factors may live in TMEM, but their layouts and lifecycles differ.

Source: distilled from "Modern GPU Programming for MLSys" (https://mlc.ai/modern-gpu-programming-for-mlsys/) and the NVIDIA Blackwell tuning/compatibility guides. Short paraphrases only; no long passages are reproduced.

I — Methodology skeleton (Interpretation)

TMEM planning resembles an "explicitly managed accumulator heap", not compiler-allocated registers:

  1. First compute the columns and alignment each object class needs;
  2. then map the logical tiles onto TLane/TCol;
  3. annotate who writes, who reads, and when each completes;
  4. arrange region reuse and the final free;
  5. ensure the epilogue's register fragments are consistent with the TMEM mapping used at write time.

A larger tile reducing register pressure does not mean the resources are free; it shifts the pressure onto the TMEM column budget and readback bandwidth.

A1 — Applications in the source (Past Application)

Case 1: dense MMA accumulator
  • tcgen05.mma writes the fp32 accumulator to TMEM.
  • After the compute phase, the full warpgroup uses tcgen05.ld to read its fragments back into registers for cast/epilogue/store.
Case 2: block-scaled MMA
  • A/B data live in SMEM.
  • Scale factors first go to SMEM, then enter the dedicated TMEM scale layout via tcgen05.cp.
  • The scale region and the accumulator region must not reuse the same mapping just because both are in TMEM.

A2 — Trigger scenarios (Future Trigger) ★

In what situations will the user need this skill?
  1. "How many TMEM columns does this tcgen05 kernel need to allocate?"
  2. "Help me plan the TMEM reuse of S/P/O in FlashAttention."
  3. "How should the completion and free ordering of tcgen05.ld/st/cp be written?"
Language signals
  • "How many TMEM columns does this tcgen05 kernel need to allocate?"
  • "Help me plan the TMEM reuse of S/P/O in FlashAttention."
  • "How should the completion and free ordering of tcgen05.ld/st/cp be written?"
Distinction from adjacent skills

Versus b200-tcgen05-mma-contract-builder: this skill focuses on TMEM resources, layout, and lifecycle; the MMA contract builder additionally decides the tile shape, cta_group, dtype, and operand placement.

Show full SKILL.md (393 more words)Show less

E — Executable steps (Execution)

Once the skill is activated, the agent must execute the following procedure:

  1. List the TMEM objects
    • Accumulators, score/prob/output tiles, scale factors, temporary regions.
    • For each object, record its dtype, logical shape, and the number simultaneously live.
  2. Compute the column budget
    • Map the objects onto 32-bit Cols; round up to the 32-column allocation unit.
    • Include the replication factors for pipeline stages, dual Q stages, or multiple consumers.
  3. Define the 2D layout
    • Give explicit TLane, TCol formulas; "so many bytes off the base" alone is not enough.
    • For cta_group::2, write out each CTA's fragment in its own TMEM separately.
  4. Define the write paths
    • MMA accumulator: tcgen05.mma.
    • Scale: tcgen05.cp from SMEM→TMEM.
    • register→TMEM: tcgen05.st.
  5. Define the readback path
    • Choose the tcgen05.ld shape/repeat; write out the fragment each warp/lane receives.
    • After readback, wait on the corresponding wait::ld before consuming the registers.
  6. Draw the lifecycle
    • allocate → producer writes → completion → consumers read/modify → completion → reuse/free.
    • Annotate every reuse edge with its barrier or wait.
  7. Check region reuse
    • Overlay only when all old consumers have completed and the new layout does not conflict.
  8. Produce the resource report
    • Total columns, peak simultaneously-live columns, alignment waste, the riskiest reuse edge, and the epilogue readback cost.
Required outputs
  1. Conclusion: the current choice/diagnosis, without vague "could be any of them" hedging.
  2. Evidence or assumptions: which come from user data, and which are hypotheses awaiting verification.
  3. Contract/table/timeline: the auditable intermediate artifacts corresponding to this skill.
  4. Minimal validation: correctness tests, boundary tests, and one falsifiable experiment.
  5. Risks and fallback: alternative paths when hardware, version, or resource conditions are not met.

B — Boundaries (Boundary) ★

Do not use when
  • The target is not Blackwell, or the kernel does not use tcgen05/TMEM.
  • The user only needs ordinary shared memory tiling.
Failure modes
  • Accessing TMEM with ld.shared/st.shared.
  • Forgetting to allocate in 32-column units, or forgetting to free.
  • Consuming registers before tcgen05.ld has completed.
  • Mixing the accumulator layout with the scale-factor layout.
Limitations
  • The exact available column count, supported load shapes, and lowering depend on the toolchain version; verify against the target compiler and the generated PTX.
  • depends-on: b200-layout-contract-auditor
  • contrasts-with: none
  • composes-with: b200-tcgen05-mma-contract-builder, b200-mbarrier-protocol-auditor, b200-flash-attention4-planner

Audit info

  • Validation passed: V1 ✓ / V2 ✓ / V3 ✓
  • Test definitions: 6 (3 should_trigger / 2 should_not_trigger / 1 edge_case)
  • Hardware validation: not performed; must be verified on a target B200
  • Distilled: 2026-06-25

© mirage-project, Apache-2.0. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

SKILL.md and 1 other file in .claude/skills/b200-tmem-lifecycle-planner of mirage-project/mirage.

  • SKILL.md
  • test-prompts.json

Open the folder on GitHubat commit f9eb70c

Compare with similar skills

B200 Tmem Lifecycle Planner 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.

B200 Tmem Lifecycle Planner compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
B200 Tmem Lifecycle Planner this skillmirage-project/mirage2.5k—~1.7kAutomated safety check: PassApache-2.0
Plannerpenpot/penpot61k—~2.7kAutomated safety check: PassMPL-2.0
Agent Plannerruvnet/ruflo74k2 repos~1.2kAutomated safety check: PassMIT
Agent Goal Plannerruvnet/ruflo74k2 repos~842Automated safety check: PassMIT
Agent Code Goal Plannerruvnet/ruflo74k2 repos~3.6kAutomated safety check: PassMIT
Travel Plannersickn33/agentic-awesome-skills47k1 repos~2kAutomated safety check: PassMIT

Similar skills

  • Planner

    penpot/penpot

    Read-only planning and architecture analysis — produce a structured implementation plan with task breakdown, acceptance criteria, sizing, and checkpoints.

    61k GitHub stars~2.7k tokensUpdated yesterday
    Agent WorkflowsAuto-check passed

  • Agent Planner

    ruvnet/ruflo

    Agent skill for planner - invoke with $agent-planner. An agent skill from ruvnet/ruflo.

    74k GitHub starsUsed in 2 repos~1.2k tokens
    Agent WorkflowsAuto-check passed

  • Agent Goal Planner

    ruvnet/ruflo

    Agent skill for goal-planner - invoke with $agent-goal-planner

    74k GitHub starsUsed in 2 repos~842 tokens
    Auto-check passed

  • Agent Code Goal Planner

    ruvnet/ruflo

    Agent skill for code-goal-planner - invoke with $agent-code-goal-planner

    74k GitHub starsUsed in 2 repos~3.6k tokens
    Agent WorkflowsAuto-check passed

  • Travel Planner

    sickn33/agentic-awesome-skills

    旅行/行程规划需求时使用:规划去某地旅行、X天X城、带老人孩子、自驾、假期安排等。产出逐日行程表、预算估算(经济/舒适/奢华三档)、交通住宿建议、景点美食清单。必须先问预算,预算未确认只输出问题清单;事实数据带来源和查询日期。

    47k GitHub starsUsed in 1 repo~2k tokens
    Auto-check passed

  • Curriculum Planner

    THU-MAIC/OpenMAIC

    Plans a multi-classroom series such as a seven-day course, clarifies the brief in rounds, gets sign-off on the full lesson list, then builds each stage in a shared folder.

    40k GitHub stars~2.8k tokensUpdated today
    EducationAuto-check passed

More from mirage-project/mirage

All 24 skills in this repo

  • V2 Perf Iteration

    mirage-project/mirage

    Runtime-V2 performance-iteration workflow. An agent skill from mirage-project/mirage.

    2.5k GitHub stars~4k tokensUpdated 3 days ago
    Auto-check passed

  • Add Mpk Task

    mirage-project/mirage

    Step-by-step guide for adding a new task implementation to Mirage Persistent Kernel (MPK).

    2.5k GitHub stars~4.5k tokensUpdated 3 days ago
    Auto-check passed

  • B200 Flash Attention4 Planner

    mirage-project/mirage

    A skill your agent uses when the user wants to design or extend a FlashAttention-style forward kernel on B200/Blackwell, involving the two MMAs QKᵀ and PV, online softmax, S/P/O in TMEM, warp roles…

    2.5k GitHub stars~1.9k tokensUpdated 3 days ago
    Auto-check passed

  • Mpk Faithful Gate

    mirage-project/mirage

    Build or run a FAITHFUL in-MPK per-task latency gate (slowCTA at the production grid + cos) for a DeepSeek-V3 MPK decode kernel or shape.

    2.5k GitHub stars~2.6k tokensUpdated 3 days ago
    Auto-check passed

  • Mpk Lever Cleanup

    mirage-project/mirage

    A skill your agent uses when a batch of env-gated (ifdef MPKDSV3 / os.environ-controlled, default-OFF) MPK optimization levers needs to be consolidated into a single clean code path for a PR…

    2.5k GitHub stars~2.2k tokensUpdated 3 days ago
    Auto-check passed

  • Test Mode

    mirage-project/mirage

    Guide for using MPK test mode to unit-test individual layers or multi-layer pipelines through the full compilation pipeline.

    2.5k GitHub stars~4.6k tokensUpdated 3 days ago
    Auto-check passed

Questions about B200 Tmem Lifecycle Planner

What does B200 Tmem Lifecycle Planner do?

A skill your agent uses when the user is using tcgen05 on Blackwell/B200 and needs to plan TMEM accumulators, block-scale factors, TMEM column allocation, the tcgen05.ld/st/cp paths, epilogue…. B200 Tmem Lifecycle Planner is an agent skill from mirage-project/mirage.ld/st/cp paths, epilogue readback, and safe deallocation.

When should I use B200 Tmem Lifecycle Planner?

B200 Tmem Lifecycle Planner fits situations like: the user is using tcgen05 on Blackwell/B200 and needs to plan TMEM accumulators; block-scale factors; TMEM column allocation; the tcgen05.ld/st/cp paths.

How do I install B200 Tmem Lifecycle Planner in Claude Code?

Run `npx skills add mirage-project/mirage --skill b200-tmem-lifecycle-planner -a claude-code`. Or copy the skill folder (.claude/skills/b200-tmem-lifecycle-planner in mirage-project/mirage) into .claude/skills/b200-tmem-lifecycle-planner in your project. Claude Code loads it when a task matches its description.

How do I install B200 Tmem Lifecycle Planner in Codex?

Run `npx skills add mirage-project/mirage --skill b200-tmem-lifecycle-planner -a codex`. Or copy the skill folder (.claude/skills/b200-tmem-lifecycle-planner in mirage-project/mirage) into .agents/skills/b200-tmem-lifecycle-planner in your project. Codex loads it when a task matches its description.

Can I use B200 Tmem Lifecycle Planner 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 mirage-project/mirage --skill b200-tmem-lifecycle-planner -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/b200-tmem-lifecycle-planner, .gemini/skills/b200-tmem-lifecycle-planner, .github/skills/b200-tmem-lifecycle-planner and .opencode/skills/b200-tmem-lifecycle-planner in your project.

What does B200 Tmem Lifecycle Planner need to run?

SKILL.md names no scripts, command-line tools or credentials: B200 Tmem Lifecycle Planner is instructions for the agent only.

Does B200 Tmem Lifecycle Planner access the network?

SKILL.md names 1 domain. As links in the text: mlc.ai. This is read from the text; nothing was executed.

Is B200 Tmem Lifecycle Planner 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 B200 Tmem Lifecycle Planner use?

B200 Tmem Lifecycle Planner is published under the Apache-2.0 licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does B200 Tmem Lifecycle Planner use?

About 1.7k tokens (SKILL.md is roughly 6.7k 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 B200 Tmem Lifecycle Planner?

Skills that share tags, products or a category with B200 Tmem Lifecycle Planner: Planner (penpot/penpot, 61k stars), Agent Planner (ruvnet/ruflo, 74k stars), Agent Goal Planner (ruvnet/ruflo, 74k stars) and Agent Code Goal Planner (ruvnet/ruflo, 74k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains B200 Tmem Lifecycle Planner?

mirage-project (a GitHub organization) maintains it in mirage-project/mirage, which has 2,545 GitHub stars. The repository holds 24 skills in this directory. The repository was last updated on October 7, 2026.

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