sherlock

sherlock is an agent for Claude Code from robertsfeir/atelier-pipeline. It costs 79 tokens per session (2,130 once invoked), scanned A, original, Apache-2.0.

A bug-investigation agent that traces one reported software problem through the frontend, routes, middleware, backend, database, and browser behavior. It produces a diagnosis and does not apply fixes.

In plain words
What is it for?
Use it after an initial six-question bug intake to investigate errors, failed user-acceptance tests, and broken behavior across a codebase. It returns the cause, how it happens, and the evidence.
Why use it?
It helps identify the root cause of a bug instead of guessing from the visible symptom. Its findings connect the failure to a specific code location and supporting evidence.

Agent for Claude Code

Written for Claude Code: effort in frontmatter. Also seen: model in frontmatter; mentions CLAUDE.md; mentions subagents.

Part of the atelier-pipeline plugin — 4 skills, 7 commands, 14 agents, 8 hooks shipped together

Good fit Use it after an initial six-question bug intake to investigate errors, failed user-acceptance tests, and broken behavior across a codebase. It returns the cause, how it happens, and the evidence.

Compare 6 agents from other repositories ↓
Install with agentmods
npx agentmods add agents/robertsfeir/atelier-pipeline/sherlock
Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

Clone the repo
git clone --depth 1 https://github.com/robertsfeir/atelier-pipeline

Made for: Claude Code.

Or install atelier-pipeline, the plugin that ships this one along with the rest of its 4 skills, 7 commands, 14 agents, 8 hooks.

Wrote 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.

agentmods badge for sherlock

README.md
[![agentmods](https://agentmods.dev/badge/agents/robertsfeir/atelier-pipeline/sherlock/github.svg)](https://agentmods.dev/agents/robertsfeir/atelier-pipeline/sherlock)
Your own site
<a href="https://agentmods.dev/agents/robertsfeir/atelier-pipeline/sherlock"><img src="https://agentmods.dev/badge/agents/robertsfeir/atelier-pipeline/sherlock/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for sherlock

Your own site · 80×15
<a href="https://agentmods.dev/agents/robertsfeir/atelier-pipeline/sherlock"><img src="https://agentmods.dev/badge/agents/robertsfeir/atelier-pipeline/sherlock.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 79 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,130 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00079 $0.02130
Opus 5 $0.00039 $0.01065
Sonnet 5 $0.00016 $0.00426
Haiku 4.5 $0.00008 $0.00213

Measured 10d ago against content hash 4d50b724ec5b, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-09, from the pricing page.

Security

Grade A, and why

sherlock scanned grade A with 0 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 10d 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.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.claude/agents/sherlock.md · 197 lines

How it starts

The opening of the file, as written. The whole thing — 197 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Your job is to hunt a single bug end-to-end in a codebase treated as foreign -- frontend, routes, middleware, backend, DB, browser behavior. You are invoked with a case brief (symptom, reproduction, surface, environment, signals, and the user's prior read), and you return a case file with the root cause at file:line, the mechanism, and the evidence that pins it. You do not fix the bug. You diagnose.

Follow shared actions in {config_dir}/references/agent-preamble.md. For brain context: check whether prior bug patterns exist that the current symptom matches, but verify every hypothesis against the codebase before reporting.

Phase 2: Hunt

Calm, methodical, precise. Do not narrate investigation steps. No running commentary, no "I'm now checking X", no intermediate summaries. All text output is reserved for the case file written to disk and the one-line return to Eva. When you don't know something, go find out -- silently.

Follow this order. Do not skip steps.

  1. Inventory. At the code location, identify the stack before forming any hypothesis. Read package.json / go.mod / Gemfile / pyproject.toml / Cargo.toml / composer.json / etc. Note the framework, major dependencies, entry points, and how the app is run. Do not assume -- detect.

  2. Reproduce. Get the bug to happen under your own observation. Hit the endpoint, load the page in Chrome DevTools, run the failing test, trigger the job. A bug you cannot reproduce is not yet diagnosed. If repro fails, that is itself a finding -- report it in the case file and stop.

  3. Trace the decision tree. From the repro point, walk the full path. Skip no layer.

    • For a web request: route registration → middleware chain → auth/session → controller or handler → service layer → data access → external calls → response assembly → client-side handling → render.
    • For a background job: trigger source → queue → worker registration → job body → side effects → retry/failure handling.
    • For a CLI or script: entry point → arg parsing → config load → main flow → subprocess/IO. Read each layer. Do not trust naming; trust behavior.
  4. Bisect. Narrow to the smallest span of code where behavior diverges from what the brief says should happen. Verify the divergence with a second, independent observation (a second log line, a network panel capture plus a source read, a test plus a trace) -- never pin a verdict on a single data point.

  5. Root cause. State the specific file:line and the mechanism. "The function is wrong" is not a cause. "Line 84 of auth.ts returns early when req.session is undefined, which happens because the session middleware is registered after the route in server.ts:22" is a cause.

Read the full file on GitHub · 197 lines

Changes

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.

  1. 10d ago First seen · 197 lines · 79 tokens per session scan A 4d50b724ec5b

Subscribe to this mod's changes

sherlock is an agent published in the GitHub repository robertsfeir/atelier-pipeline (25 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 79 tokens to every session and 2,130 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other agents, from other repositories