Pyroscope

What it does

When your agent uses it

• Adding profiling to a service
• Deploying Alloy as a cluster-wide eBPF profiler
• Hunting CPU / memory hotspots from a flame graph
• Comparing two profiles to find a regression

Example prompts

• “find what”
• “flame graph this app”
• “continuous profiling”
• “/pyroscope”

Requirements

• Python 3
• A credential in GRAFANA_API_KEY

Workflow steps

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

1. 1Instrument an app with the SDK (representative: Python)
2. 2Cluster-wide eBPF profiling with Alloy
3. 3Query with ProfileQL

What it can do on your machine

Read from SKILL.md and the folder at commit 1ccacf2. 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:

  • curl
  • jq
  • pip

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

• Network

  Hosts in commands or code, which the agent is likely to contact:

  • profiles-prod-xxx.grafana.net

  Also links to:

  • grafana.com

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

• Credentials

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

  • GRAFANA_API_KEY

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

Context cost

Pyroscope loads about 1.2k tokens when it runs, and up to ~2.7k if it reads all its reference files. Until then it costs about 223 tokens; SKILL.md has 208 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it

~223

When it runs · the whole SKILL.md, loaded when a task matches

~1.2k

With references · SKILL.md plus every file in references/, read only if the agent opens them

~2.7k

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

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 grafana/skills at commit 1ccacf2, republished under its Apache-2.0 licence (© grafana). 208 words, ~1,189 tokens.

name

pyroscope

description

Continuously profile applications with Grafana Pyroscope and read the result as flame graphs. Covers three instrumentation paths — language SDK push (Go / Java / Python / Ruby / Node / .NET / Rust), Alloy eBPF auto-instrumentation (no code change, requires kernel 5.8+ with BTF), and SDK → Alloy receiver — plus ProfileQL queries, profile types (CPU / memory / allocations / goroutines / mutex), Grafana Cloud Profiles endpoint, and Span Profiles trace-to-profile linking. Use when adding profiling to a service, deploying Alloy as a cluster-wide eBPF profiler, hunting CPU / memory hotspots from a flame graph, comparing two profiles to find a regression, or correlating a slow Tempo trace to its profile — even when the user says "find what's burning CPU", "flame graph this app", "continuous profiling", "heap hotspots", or "why is allocation so high" without naming Pyroscope.

license

Apache-2.0

pip install pyroscope-io==1.0.11

import pyroscope, os
pyroscope.configure(
    application_name="my.python.app",
    server_address="http://pyroscope:4040",
    sample_rate=100, oncpu=True,
    tags={"region": os.getenv("REGION"), "env": "prod"},
)
# Dynamic tag for a hot path
with pyroscope.tag_wrapper({"controller": "slow_controller"}):
    slow_code()

# Verify the app is pushing — Pyroscope ingests samples in a few seconds
curl -s http://pyroscope:4040/ready                                    # → "ready"
curl -s http://pyroscope:4040/api/v1/labels | jq '.data | index("service_name")'  # → not null

# Verify the service shows up in Grafana → Explore → Profiles → service dropdown.

# config.alloy — full block in references/ebpf-and-query.md
pyroscope.ebpf "local_pods" {
  forward_to       = [pyroscope.write.cloud.receiver]
  targets          = discovery.relabel.local_pods.output
  sample_rate      = 97
  collect_interval = "15s"
}
pyroscope.write "cloud" {
  endpoint {
    url = "https://profiles-prod-xxx.grafana.net"
    basic_auth { username = sys.env("PYROSCOPE_USER")
                 password = sys.env("GRAFANA_API_KEY") }
  }
}

# 1. Reload Alloy
curl -X POST http://localhost:12345/-/reload

# 2. Verify the eBPF component is healthy
curl -s http://localhost:12345/api/v0/web/components \
  | jq '.[] | select(.id|contains("pyroscope.ebpf")) | {id,health:.health.state}'
# Expect: health.state == "healthy"

# 3. Verify profiles arriving in Pyroscope
#    Grafana → Explore → Profiles datasource → query:
#      {namespace="default", __profile_type__="process_cpu:cpu:nanoseconds:cpu:nanoseconds"}
#    Expect flame graph to render with frames from the target pods.

{service_name="myapp", env="prod",
 __profile_type__="process_cpu:cpu:nanoseconds:cpu:nanoseconds"}

© grafana, 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 2 other files (references) in skills/grafana-lgtm/pyroscope of grafana/skills.

• SKILL.md
• references/ebpf-and-query.md
• references/sdks.md

Open the folder on GitHubat commit 1ccacf2