Agent skill

Alloc Profile

by ClickHouse in ClickHouse/ClickHouse

Analyze a jemalloc (or other) allocation profile in collapsed stack format.

Apache-2.0Auto-check passedDatabases

Install Alloc Profile

skills CLI
$ npx skills add ClickHouse/ClickHouse --skill alloc-profile -a claude-code

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

GitHub CLI
$ gh skill install ClickHouse/ClickHouse alloc-profile --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/ClickHouse/ClickHouse.git skills-src && mkdir -p .claude/skills && cp -r skills-src/.claude/skills/alloc-profile .claude/skills/alloc-profile && 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
alloc-profile
GitHub stars
50k
Token cost
~4.3k tokens
SKILL.md length
1,086 words
Files
1
Skills in repo
24
Repo updated
First seen
Licence
Apache-2.0

At a glance

Analyze a jemalloc (or other) allocation profile in collapsed stack format.

  • Works in 4 steps: Locate the profile file → Parallel initial analysis → Synthesize results and perform subsystem… → …
  • The user wants to analyze memory allocations
  • SKILL.md covers Arguments, Step 1 — Locate the profile file, Step 2 — Parallel initial… and Step 3 — Synthesize results…, plus 3 more sections
  • Calls python3

What it does

Alloc Profile is an agent skill from ClickHouse/ClickHouse. Analyze a jemalloc (or other) allocation profile in collapsed stack format. Use when the user wants to analyze memory allocations, find top allocators, or understand memory usage patterns from a .collapsed profile file.

Its SKILL.md is about 4.3k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Databases. It works with Bash and Python. The repository describes itself as: ClickHouse® is a real-time analytics database management system. The licence is Apache-2.0.

When your agent uses it

  • The user wants to analyze memory allocations
  • Find top allocators
  • Understand memory usage patterns from a .collapsed profile file

Example prompts

  • “/alloc-profile”

Requirements

  • Python 3
  • Pre-approved tools (allowed-tools): Task, Bash(ls:*), Bash(find:*), Bash(wc:*), Bash(python3:*)

Workflow steps

4 steps, taken from the step headings in SKILL.md.

  1. Locate the profile file
  2. Parallel initial analysis
  3. Synthesize results and perform subsystem grouping
  4. Offer drill-down options

What it can do on your machine

Read from SKILL.md and the folder at commit cd023af. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves these tools, so the agent can use them without asking each time:

    • Task
    • Bash(ls:*)
    • Bash(find:*)
    • Bash(wc:*)
    • Bash(python3:*)

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • python3

    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 no API keys, tokens, secrets or passwords.

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

Context cost

Alloc Profile loads about 4.3k tokens when it runs. Until then it costs about 58 tokens; SKILL.md has 1,086 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~58
When it runs · the whole SKILL.md, loaded when a task matches
~4.3k

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 ClickHouse/ClickHouse at commit cd023af, republished under its Apache-2.0 licence (© ClickHouse). 1,086 words, ~4,317 tokens.

Download SKILL.mdSave it as .claude/skills/alloc-profile/SKILL.md (or your agent's skills folder).
name
alloc-profile
description
Analyze a jemalloc (or other) allocation profile in collapsed stack format. Use when the user wants to analyze memory allocations, find top allocators, or understand memory usage patterns from a .collapsed profile file.
allowed-tools
Task, Bash(ls:*), Bash(find:*), Bash(wc:*), Bash(python3:*)
argument-hint
path-to-profile.collapsed
disable-model-invocation
false

Allocation Profile Analysis Skill

Analyze an allocation profile file in collapsed stack format (as produced by jemalloc, async-profiler, perf). Each line has the form:

frame1;frame2;...;frameN VALUE

where VALUE is the number of bytes (or samples, depending on the profiler) attributed to that stack trace.

Arguments

  • $0 (optional): Path to the .collapsed file. If not provided, search for .collapsed files in the current directory and ask the user to choose.

Step 1 — Locate the profile file

Use Task tool with subagent_type=Bash to locate the file:

If $ARGUMENTS is provided, use it directly. Otherwise, run:

bash
find . -maxdepth 3 -name "*.collapsed" -o -name "*.folded" | sort -t_ -k1,1

Report the candidates to the user and ask with AskUserQuestion:

  • Question: "Which profile file do you want to analyze?"
  • Options: one per found file (show filename and size), plus "Other — enter path manually"

Once the file path is known, pass it to all subsequent steps.

Step 2 — Parallel initial analysis

Launch the following three Task agents IN PARALLEL (single message, three tool calls) all with run_in_background: true. Then call TaskOutput for ALL three agents (also in parallel, single message) before proceeding to Step 3. Do NOT start Step 3 until every agent has finished.

Fallback: If any agent fails (e.g., reports lacking Bash permission), re-run its Python script directly using the Bash tool in the main context.

Agent A — Summary statistics (subagent_type=Bash)

Run this Python script to compute summary statistics:

python
python3 - <<'EOF'
import sys, os, re
filepath = "PATH_TO_FILE"  # substituted by skill
lines = open(filepath).read().splitlines()
traces = []
for line in lines:
    line = line.strip()
    if not line:
        continue
    parts = line.rsplit(' ', 1)
    if len(parts) != 2:
        continue
    try:
        traces.append((int(parts[1]), parts[0]))
    except ValueError:
        continue

total = sum(v for v, _ in traces)
traces.sort(reverse=True)

# Noise filters — keep in sync with Agent C
JEMALLOC_PREFIXES = (
    "prof_backtrace", "prof_alloc_prep", "prof_tctx", "prof_",
    "imalloc", "ialloc", "irallocx", "imallocx",
    "arena_malloc", "arena_palloc", "arena_ralloc", "arena_",
    "tcache_alloc", "tcache_",
    "large_malloc", "large_palloc",
    "chunk_alloc", "huge_malloc", "huge_palloc",
    "je_malloc", "je_calloc", "je_realloc", "je_rallocx", "je_mallocx",
    "je_posix_memalign", "je_aligned_alloc",
    "malloc_default", "calloc",
)
ALLOC_SUBSTRINGS = (
    "operator new", "operator new[]",
    "__libcpp_operator_new",
    "__libc_malloc", "__libc_calloc", "_int_malloc",
    "posix_memalign", "aligned_alloc",
    "do_rallocx", "do_mallocx",
    "mi_malloc", "mi_calloc",
    "__cxx_global_var_init", "__cxa_thread_atexit_impl",
    "DB::Memory<", "Memory::newImpl", "Allocator<false", "Allocator<true",
    "allocNoTrack",
    "PODArrayBase::realloc", "PODArrayBase::alloc",
    "CRYPTO_malloc",
    "std::__detail::_Hash_node", "std::_Rb_tree",
    "std::vector<", "std::string::",
    # STL and PODArray wrappers — noise for leaf analysis
    "std::__1::",
    "DB::PODArrayBase",
)

def is_noise(frame):
    return (any(frame.startswith(p) for p in JEMALLOC_PREFIXES) or
            any(s in frame for s in ALLOC_SUBSTRINGS))

def shorten(frame):
    return re.sub(r'<[^>]{40,}>', '<...>', frame)

print(f"=== SUMMARY ===")
print(f"File: {filepath}")
print(f"Total allocated: {total:,} bytes  ({total/1024/1024:.2f} MB)  ({total/1024/1024/1024:.3f} GB)")
print(f"Unique stack traces: {len(traces)}")
print()
print("=== TOP 25 STACK TRACES ===")
for i, (v, stack) in enumerate(traces[:25], 1):
    frames = [f for f in stack.split(';') if f]
    meaningful = [f for f in frames if not is_noise(f)]
    tail_frames = meaningful[-4:] if meaningful else frames[-4:]
    tail = ' <- '.join(shorten(f) for f in reversed(tail_frames))
    print(f"{i:>3}. {v/1024/1024:>8.2f} MB  ({100*v/total:>5.1f}%)  {tail[:120]}")
print()
print("=== FULL STACKS FOR TOP 10 ===")
for i, (v, stack) in enumerate(traces[:10], 1):
    frames = [f for f in stack.split(';') if f]
    print(f"\n--- #{i}: {v/1024/1024:.2f} MB ({100*v/total:.1f}%) ---")
    for depth, frame in enumerate(reversed(frames), 1):
        noise_mark = "  [noise]" if is_noise(frame) else ""
        print(f"  [{depth:>2}] {shorten(frame)}{noise_mark}")
EOF
Agent B — Outermost meaningful frame aggregation (subagent_type=Bash)

Run this Python script to aggregate by the outermost (shallowest) meaningful frame — the operation that initiated the allocation. This answers "why did this allocation happen?" (e.g., loading data parts, executing a query, loading a dictionary), complementing Agent C which answers "what code allocated?":

python
python3 - <<'EOF'
import sys, re
from collections import defaultdict
filepath = "PATH_TO_FILE"  # substituted by skill

lines = open(filepath).read().splitlines()
traces = []
for line in lines:
    line = line.strip()
    if not line:
        continue
    parts = line.rsplit(' ', 1)
    if len(parts) != 2:
        continue
    try:
        traces.append((int(parts[1]), parts[0]))
    except ValueError:
        continue

total = sum(v for v, _ in traces)

# Frames to skip when looking for the outermost meaningful frame:
# thread pool scaffolding, libc entry points, raw addresses, lambda wrappers
SKIP_OUTER = (
    "0000", "_start", "__libc_start", "__GI___clone",
    "start_thread", "clone3",
    "ThreadPoolImpl", "ThreadFromGlobalPool",
    "std::__1::__function", "std::__1::__invoke",
    "decltype", "void std::__1::__function",
    "std::__1::__packaged_task_function",
    "DB::ThreadPool", "DB::GlobalThreadPool",
    "DB::threadFunction",
    "BaseDaemon", "SignalListener",
    "Poco::ThreadImpl::runnableEntry",
    "Poco::PooledThread::run",
    "main",
    "DB::Server::run",
    "Poco::Util::Application::run",
)

def is_skip_outer(frame):
    return any(frame.startswith(p) for p in SKIP_OUTER) or frame.startswith("(")

def shorten(frame):
    # Collapse long templates, preserve (anonymous namespace), strip args
    s = re.sub(r'<[^>]{40,}>', '<...>', frame)
    s = s.replace('(anonymous namespace)', '{anon}')
    s = re.sub(r'\(.*', '', s)
    s = s.replace('{anon}', '(anonymous namespace)')
    return s[:120]

by_outer = defaultdict(int)

for v, stack in traces:
    frames = [f for f in stack.split(';') if f]
    outer = None
    for f in frames:
        if not f or is_skip_outer(f):
            continue
        outer = f
        break
    if outer is None:
        outer = frames[0] if frames else "(unknown)"
    by_outer[shorten(outer)] += v

print("=== TOP 25 OUTERMOST MEANINGFUL FRAMES (operation that initiated allocation) ===")
for fn, v in sorted(by_outer.items(), key=lambda x: -x[1])[:25]:
    mb = v / 1024 / 1024
    pct = 100 * v / total
    bar = "\u2588" * int(pct / 2)
    print(f"  {mb:>10.2f} MB  {pct:>5.1f}%  {bar:<20}  {fn}")
EOF
Agent C — Leaf (allocating) function aggregation (subagent_type=Bash)

Run this Python script to aggregate by the deepest (innermost) frame — the actual allocation call:

python
python3 - <<'EOF'
import sys, re
from collections import defaultdict
filepath = "PATH_TO_FILE"  # substituted by skill

lines = open(filepath).read().splitlines()
traces = []
for line in lines:
    line = line.strip()
    if not line:
        continue
    parts = line.rsplit(' ', 1)
    if len(parts) != 2:
        continue
    try:
        traces.append((int(parts[1]), parts[0]))
    except ValueError:
        continue

total = sum(v for v, _ in traces)

# Aggregate by last meaningful frame (the allocating function)
by_leaf = defaultdict(int)
by_caller = defaultdict(int)  # caller of the leaf

# jemalloc profiling infrastructure — always at the bottom of every stack
JEMALLOC_PREFIXES = (
    "prof_backtrace", "prof_alloc_prep", "prof_tctx", "prof_",
    "imalloc", "ialloc", "irallocx", "imallocx",
    "arena_malloc", "arena_palloc", "arena_ralloc", "arena_",
    "tcache_alloc", "tcache_",
    "large_malloc", "large_palloc",
    "chunk_alloc", "huge_malloc", "huge_palloc",
    "je_malloc", "je_calloc", "je_realloc", "je_rallocx", "je_mallocx",
    "je_posix_memalign", "je_aligned_alloc",
    "malloc_default", "calloc",
)
# libc / C++ allocator wrappers that add no information
ALLOC_SUBSTRINGS = (
    "operator new", "operator new[]",
    "__libcpp_operator_new",
    "__libc_malloc", "__libc_calloc", "_int_malloc",
    "posix_memalign", "aligned_alloc",
    "do_rallocx", "do_mallocx",
    "mi_malloc", "mi_calloc",
    # C++ static/thread-local initialization wrappers
    "__cxx_global_var_init", "__cxa_thread_atexit_impl",
    # ClickHouse allocator wrappers — informative only as callers, not as leaf
    "DB::Memory<", "Memory::newImpl", "Allocator<false", "Allocator<true",
    "allocNoTrack",
    "PODArrayBase::realloc", "PODArrayBase::alloc",
    # Third-party allocators
    "CRYPTO_malloc",
    # STL internals
    "std::__detail::_Hash_node", "std::_Rb_tree",
    "std::vector<", "std::string::",
    # STL and PODArray wrappers — noise for leaf analysis
    "std::__1::",
    "DB::PODArrayBase",
)

def is_noise(frame):
    return (any(frame.startswith(p) for p in JEMALLOC_PREFIXES) or
            any(s in frame for s in ALLOC_SUBSTRINGS))

def meaningful_leaf(frames):
    # Walk from innermost (last) frame upward, skipping allocator/profiling noise.
    # In jemalloc collapsed format frames are outermost-first, so the bottom of
    # the stack (profiling infra + raw allocators) is at the end of the list.
    for f in reversed(frames):
        if f and not is_noise(f):
            return f
    return frames[-1] if frames else "(unknown)"

def meaningful_caller(frames):
    """Second non-noise frame from the bottom."""
    found_leaf = False
    for f in reversed(frames):
        if f and not is_noise(f):
            if found_leaf:
                return f
            found_leaf = True
    return None

def shorten(frame):
    s = re.sub(r'<[^>]{40,}>', '<...>', frame)
    s = s.replace('(anonymous namespace)', '{anon}')
    s = re.sub(r'\(.*', '', s)
    s = s.replace('{anon}', '(anonymous namespace)')
    return s[:120]

for v, stack in traces:
    frames = [f for f in stack.split(';') if f]
    leaf = meaningful_leaf(frames)
    by_leaf[shorten(leaf)] += v
    caller = meaningful_caller(frames)
    if caller:
        by_caller[shorten(caller)] += v

print("=== TOP 25 ALLOCATING FUNCTIONS (first non-trivial frame from bottom) ===")
for label, bucket in [("Leaf (allocator call site)", by_leaf),
                       ("Caller of leaf", by_caller)]:
    print(f"\n--- {label} ---")
    for fn, v in sorted(bucket.items(), key=lambda x: -x[1])[:25]:
        mb = v / 1024 / 1024
        pct = 100 * v / total
        print(f"  {mb:>8.2f} MB  {pct:>5.1f}%  {fn}")
EOF

Step 3 — Synthesize results and perform subsystem grouping

MANDATORY: All three agents from Step 2 must have completed (TaskOutput returned) before this step.

Using the outputs from Agent A (top stacks with full traces), Agent B (outermost frame — the initiating operation), and Agent C (leaf function — the allocating code), you (the main LLM) produce a structured report. Agent B gives you the "why" (what operation triggered allocations) and Agent C gives you the "how" (what code did the allocating). Combined with Agent A's full stacks, you can semantically group allocations into subsystems — e.g., AggregatedDataVariants::init called from HashedDictionary::loadData is "Dictionary Loading", not "Aggregation"; Arena::addMemoryChunk inside a merge pipeline is "Merges", not "Arena".

Your report should include:

  1. Summary statistics (total, trace count)
  2. Top allocators table (top 15 stack traces with readable short description)
  3. Subsystem breakdown with bar chart — group every top-25 trace (and the leaf function data) into semantic subsystems based on the full call path context. Use categories like: Part Loading, Dictionary Loading, Query Execution, Backup & Restore, Merges & Mutations, File Cache, IO Buffers, Replication, System Logs, etc. Show an ASCII bar chart. Traces that don't fit neatly into a category go into "(other)".
  4. Top 3–5 actionable findings — e.g.:
    • Which subsystem unexpectedly dominates
    • Any single allocation that is disproportionately large (>5% of total)
    • Repeated patterns (e.g., many system log types each reserving large buffers)
    • Signs of fragmentation or excessive reallocation (do_rallocx / PODArray::realloc heavy)
  5. Follow-up drill-down questions the user may want to investigate
Show full SKILL.md (548 more words)Show less

Step 4 — Offer drill-down options

After presenting the summary, use AskUserQuestion:

Question: "What would you like to do next?"

  • Option 1: "Drill into a specific subsystem" Description: "Show all stack traces for a chosen component (e.g., MergeTree, SystemLog)" → Ask which subsystem with a follow-up AskUserQuestion → Launch Task (subagent_type=Bash) in the background (run_in_background: true) with a Python script that:

    • Filters all traces whose stack contains any keyword matching the chosen subsystem
    • Sorts by value descending
    • Prints each trace as: MB (pct%) | frame1 ← frame2 ← ... ← frameN
    • Also prints the full call stack for the top 5 matches
    • Prints a sub-total for the subsystem → Use TaskOutput to wait, then pass output to a general-purpose Task agent for a concise summary
  • Option 2: "Show full stacks for top N traces" Description: "Print complete call stacks for the largest N allocations" → Ask N with a follow-up AskUserQuestion (suggest 10 as default) → Launch Task (subagent_type=Bash) in the background with a Python script that:

    • Parses the file, sorts by value, takes top N
    • For each: prints rank, MB, %, and the full reversed call stack with depth indices → Use TaskOutput to wait, then pass output to a general-purpose Task agent for a concise narrative summary
  • Option 3: "Search for a keyword in stacks" Description: "Filter traces containing a specific function or class name" → Ask for the keyword via AskUserQuestion → Launch two Task agents in parallel (run_in_background: true):

    • Agent X (subagent_type=Bash): filter and aggregate all matching traces — sum total, count, top 20 by size, full stacks for top 5
    • Agent Y (subagent_type=Bash): find related keywords by scanning all frames containing the keyword and extracting their neighboring frames (co-occurring functions), to suggest related call paths → Use TaskOutput (both) then pass combined output to a general-purpose Task agent for synthesis
  • Option 4: "Generate flamegraph SVG" Description: "Render an SVG flamegraph using flamegraph.pl (must be installed)" → Launch Task (subagent_type=Bash) in the background:

    bash
    flamegraph.pl --title "Allocation Profile" --countname bytes --width 1800 \
      PATH_TO_FILE > /tmp/alloc_flamegraph.svg

    → Use TaskOutput to wait for completion → Report the output path /tmp/alloc_flamegraph.svg and remind user to open it in a browser

  • Option 5: "Done" Description: "Exit without further analysis"

IMPORTANT: For every drill-down option (1–4):

  • Always run the analysis inside a Task subagent — never process the file in the main context
  • Always run the Bash analysis task in the background with run_in_background: true and wait with TaskOutput
  • Always pass raw output through a general-purpose Task agent for a concise, human-readable summary before showing it to the user

Repeat drill-down (return to the AskUserQuestion) until user selects "Done".

Notes

  • Values in collapsed format are live (in-use) bytes — jemalloc heap profiles track allocations minus deallocations, so values reflect currently live memory at the time of the dump
  • High values directly indicate live memory pressure at those call sites
  • Frames are listed outermost (thread root) first, innermost (allocator) last — the analysis scripts reverse this for readability
  • Symbol names may be mangled if the binary lacks debug info; use jeprof --demangle or pipe through c++filt
  • ALWAYS use Task subagents for all analysis steps — profile files can be hundreds of MB and must not be read into the main context
  • All Python analysis scripts are self-contained and can be run directly with python3 -

Examples

  • /alloc-profile — Find .collapsed files and prompt for selection
  • /alloc-profile jemalloc-profile-2026-02-19T13-08-59-825Z.collapsed — Analyze a specific file
  • /alloc-profile /tmp/prod-heap-dump.collapsed — Analyze an absolute path

© ClickHouse, 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

Just SKILL.md in .claude/skills/alloc-profile of ClickHouse/ClickHouse.

Open the folder on GitHubat commit cd023af

Compare with similar skills

Alloc Profile 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.

Alloc Profile compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Alloc Profile this skillClickHouse/ClickHouse50k—~4.3kAutomated safety check: PassApache-2.0
Jeecg Onlformjeecgboot/skills239—~7.7kAutomated safety check: PassApache-2.0
CLI DeveloperJeffallan/claude-skills12k2 repos~1.2kAutomated safety check: PassMIT
Create Experkfly/ex-skill2.5k—~3.7kAutomated safety check: NotesMIT
Hugging Face API Tool Builderhuggingface/skills11k5 repos~1.5kAutomated safety check: PassApache-2.0
Markdown Exporterbowenliang123/markdown-exporter2711 repos~5.3kAutomated safety check: PassApache-2.0

Similar skills

  • Jeecg Onlform

    jeecgboot/skills

    JeecgBoot Online表单(cgform)全生命周期管理——通过API自动创建/编辑数据库表和表单配置, 支持单表、主子表、树表,26种控件类型,以及JS/Java/SQL增强、权限配置、数据CRUD、积木报表集成。

    239 GitHub stars~7.7k tokensUpdated 20 days ago
    DatabasesAuto-check passed
  • CLI Developer

    Jeffallan/claude-skills

    Walks through designing, building and polishing a command-line tool: user workflow and command hierarchy, implementation in commander, click, typer or cobra, completions and cross-platform testing.

    12k GitHub starsUsed in 2 repos~1.2k tokens
    DevelopmentAuto-check passed
  • Create Ex

    perkfly/ex-skill

    Distill an ex-girlfriend into an AI Skill. An agent skill from perkfly/ex-skill.

    2.5k GitHub stars~3.7k tokensUpdated 6 mo ago
    Productivity & AutomationAuto-check: notes
  • Official

    Builds reusable command line scripts that fetch, enrich or process data from the Hugging Face API, aimed at chained, repeated or automated tasks.

    11k GitHub starsUsed in 5 repos~1.5k tokens
    AI & LLM EngineeringAuto-check passed
  • Markdown Exporter

    bowenliang123/markdown-exporter

    Convert Markdown text to DOCX, PPTX, XLSX, PDF, PNG, SVG, HTML, IPYNB, MD, CSV, JSON, JSONL, XML files, and extract code blocks in Markdown to Python, Bash,JS and etc files.

    271 GitHub starsUsed in 1 repo~5.3k tokens
    Documents & OfficeAuto-check passed
  • K8e Sandbox

    xiaods/k8e

    Run a goal end to end inside an isolated K8E sandbox pod (gVisor / Kata / Firecracker) instead of on the host: exec bash / Python / Node / TypeScript, install packages, move files in and out, reuse…

    500 GitHub stars~6k tokensUpdated 9 days ago
    Backend & APIsAuto-check passed

More from ClickHouse/ClickHouse

All 24 skills in this repo
  • Keeper Stress Analysis

    ClickHouse/ClickHouse

    Analyze ClickHouse Keeper stress-test results from play.clickhouse.com / keeperstresstests data warehouse.

    50k GitHub stars~4.7k tokensUpdated today
    Auto-check passed
  • Perf Comparison

    ClickHouse/ClickHouse

    Evaluate ClickHouse performance test results from existing CI/dashboard data or local perf.py runs.

    50k GitHub stars~3.9k tokensUpdated today
    Auto-check: notes
  • Patch Release Check

    ClickHouse/ClickHouse

    Check whether ClickHouse's supported versions (last 3 majors + latest LTS) have recent stable patch releases, diagnose why the scheduled AutoReleases pipeline failed, and identify which releases…

    50k GitHub stars~4k tokensUpdated today
    Auto-check: notes
  • Bisect

    ClickHouse/ClickHouse

    Bisect a ClickHouse regression using pre-built master binaries from CI.

    50k GitHub stars~1.4k tokensUpdated today
    Auto-check passed
  • Clickhouse PR Description

    ClickHouse/ClickHouse

    Generate PR descriptions for ClickHouse/ClickHouse that match maintainer expectations.

    50k GitHub stars~1.9k tokensUpdated today
    Auto-check passed
  • Codex Review

    ClickHouse/ClickHouse

    Run a local pre-push AI code review of the current branch with the OpenAI codex CLI, the same way ClickHouse CI does, and optionally iterate fix and re-review until clean.

    50k GitHub stars~1.8k tokensUpdated today
    Auto-check: notes

Works with

Categories

Questions about Alloc Profile

What does Alloc Profile do?

Analyze a jemalloc (or other) allocation profile in collapsed stack format. Alloc Profile is an agent skill from ClickHouse/ClickHouse. Analyze a jemalloc (or other) allocation profile in collapsed stack format.

When should I use Alloc Profile?

Alloc Profile fits situations like: the user wants to analyze memory allocations; find top allocators; understand memory usage patterns from a .collapsed profile file.

How do I install Alloc Profile in Claude Code?

Run `npx skills add ClickHouse/ClickHouse --skill alloc-profile -a claude-code`. Or copy the skill folder (.claude/skills/alloc-profile in ClickHouse/ClickHouse) into .claude/skills/alloc-profile in your project. Claude Code loads it when a task matches its description.

How do I install Alloc Profile in Codex?

Run `npx skills add ClickHouse/ClickHouse --skill alloc-profile -a codex`. Or copy the skill folder (.claude/skills/alloc-profile in ClickHouse/ClickHouse) into .agents/skills/alloc-profile in your project. Codex loads it when a task matches its description.

Can I use Alloc Profile 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 ClickHouse/ClickHouse --skill alloc-profile -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/alloc-profile, .gemini/skills/alloc-profile, .github/skills/alloc-profile and .opencode/skills/alloc-profile in your project.

What does Alloc Profile need to run?

Going by SKILL.md and its folder, Alloc Profile needs the command-line tools its instructions call (python3). Our summary lists: Python 3. Its frontmatter pre-approves these tools: Task, Bash(ls:*), Bash(find:*), Bash(wc:*), Bash(python3:*).

Does Alloc Profile 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 Alloc Profile 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 Alloc Profile use?

Alloc Profile 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 Alloc Profile use?

About 4.3k tokens (SKILL.md is roughly 17k 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 Alloc Profile?

Skills that share tags, products or a category with Alloc Profile: Jeecg Onlform (jeecgboot/skills, 239 stars), CLI Developer (Jeffallan/claude-skills, 12k stars), Create Ex (perkfly/ex-skill, 2.5k stars) and Hugging Face API Tool Builder (huggingface/skills, 11k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Alloc Profile?

ClickHouse (a GitHub organization) maintains it in ClickHouse/ClickHouse, which has 50,288 GitHub stars. The repository holds 24 skills in this directory. The repository was last updated on October 8, 2026.

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