mcpbeat

Investigate

oliver-kriska/investigate

Investigate bugs and errors in Elixir/Phoenix — root-cause analysis for crashes, exceptions, stack traces, test failures. Use --parallel for deep 4-track investigation.

2k tokens
context cost
the whole folder, loaded on every use
4
files
instructions only
0
copies elsewhere
how many repositories repackaged it
514
stars on the repo
on the repository, not the skill itself

Install

one command, takes just this skill from the repository
npx skills add https://github.com/oliver-kriska/claude-elixir-phoenix --skill investigate

What comes with it

4 171 bytes besides the instruction
references/debug-commands.md
references/error-patterns.md
references/investigation-template.md

What it tells the agent to use

found in the instruction text
Task spawns other agents

The instruction itself

17 sections, as written by the author

Investigate Bug

Investigate bugs using the Ralph Wiggum approach: check the

obvious, read errors literally.

Usage

/phx:investigate Users can't log in after password reset
/phx:investigate FunctionClauseError in UserController.show
/phx:investigate Complex auth bug --parallel

Arguments

$ARGUMENTS = Bug description or error message. Add --parallel

for deep 4-track investigation.

Mode Selection

Use parallel mode (spawn deep-bug-investigator) when:

bug mentions 3+ modules, spans multiple contexts, is intermittent

or involves concurrency, or user says --parallel/deep.

Before spawning it, determine the effective maximum nesting depth. Use an

explicit positive-integer CLAUDE_CODE_MAX_SUBAGENT_SPAWN_DEPTH value first;

when it is unset, inspect claude --version (the default is 1 in 2.1.217–2.1.218

and 3 in 2.1.219+). If the version is unavailable, conservatively use 1. At

depth 3+, use the deep-bug-investigator orchestrator. At depth 1 or 2, keep

orchestration in this main session: spawn the four

focused tracks (reproduction, root cause, impact, fix strategy) directly in

one parallel batch, wait for all four, then synthesize their evidence. Never

spawn an orchestrator that cannot delegate.

Otherwise: Run the sequential workflow below.

Avoid confirmatory subagents: Do NOT spawn parallel subagents

to "verify" findings you already identified in the main context.

If Step 3-4 already identified the root cause with high confidence,

present it directly — don't spend ~80K tokens on 4 subagents to

confirm what's already obvious (confirmed waste: session c135330a).

Iron Laws

  • Read the error message literally first — Most bugs tell you exactly what's wrong; resist the urge to theorize before reading what the system is saying
  • Check the obvious before going deep — Compile errors, missing migrations, atom/string mismatches explain 80% of bugs; exhausting the Ralph Wiggum checklist saves hours
  • Check changeset errors before UI debugging — Silent form saves are almost always {:error, changeset} with validation failures, not viewport or JS issues
  • Consult compound docs before investigating fresh — A previously solved problem saves the entire investigation cycle; always search .claude/solutions/ first
  • NEVER guess at a fix before reproducing — Reproduce first, then identify root cause, then fix. Skipping steps causes wrong fixes
  • DO NOT apply a fix without confirming root cause — Verify your hypothesis with evidence (logs, tests, IO.inspect) before changing code

Investigation Workflow

Step 0: Consult Compound Docs

Search .claude/solutions/ for relevant keywords using Grep.

If matching solution exists, present it and ask: "Apply this

fix, or investigate fresh?"

Step 0a: Runtime Auto-Capture (Tidewave -- PRIMARY when available)

If Tidewave MCP is detected, **start here instead of asking

the user to paste errors**. Auto-capture runtime context:

  • mcp__tidewave__get_logs level: :error -- capture recent errors
  • Parse stacktraces, correlate with source via

mcp__tidewave__get_source_location

  • For data bugs: mcp__tidewave__execute_sql_query to inspect state
  • For logic bugs: mcp__tidewave__project_eval to test hypotheses
  • For UI bugs: mcp__tidewave__get_source_location with component name

Present pre-populated context to the user:

> Auto-captured from runtime:

>

> - Error: {parsed error from logs}

> - Location: {file:line from get_source_location}

>

> Investigating this. Correct if wrong.

This eliminates copy-pasting errors between app and agent.

If Tidewave NOT available: Fall through to Step 1.

Step 1: Sanity Checks

Run mix compile --warnings-as-errors 2>&1 | head -50, then mix ecto.migrate.

Step 2: Reproduce

Run mix test test/path_test.exs --trace. Then read the last 200 lines of log/dev.log and search for "error" or "exception" patterns.

Step 3: Read Error LITERALLY

Parse the error message — check ${CLAUDE_SKILL_DIR}/references/error-patterns.md.

Step 4: Check the Obvious (Ralph Wiggum Checklist)

File saved? Atom vs string? Data preloaded? Pattern match

correct? Nil? Return value? Server restarted?

LiveView form saves silently failing? Check changeset errors

FIRST — not viewport, click mechanics, or JS. A missing

hidden_input for a required embedded field causes `{:error,

changeset}` with no visible UI feedback.

Step 5: IO.inspect / Tidewave project_eval

Step 6: Identify Root Cause

Find what's actually happening vs what should happen.

Step 7: Hand Off

Present root cause + evidence. Then route by fix size:

  • Small, contained fix → offer to apply directly or via /phx:quick
  • Multi-file or risky fix → suggest /phx:plan {root cause summary} so

the fix gets task structure and review

  • Non-obvious root cause → after the fix lands, suggest /phx:compound

Autonomous Iteration

Use /ralph-loop:ralph-loop for autonomous debugging with

clear completion criteria and --max-iterations.

References

  • ${CLAUDE_SKILL_DIR}/references/error-patterns.md — Common errors and checklist
  • ${CLAUDE_SKILL_DIR}/references/investigation-template.md — Output format
  • ${CLAUDE_SKILL_DIR}/references/debug-commands.md — Debug commands and common fixes

How to use it

Copy the folder

Take oliver-kriska/investigate from the repository into ~/.claude/skills for personal use, or into .claude/skills inside a project.

Check the name does not clash

The agent identifies a skill by the name field in its header. Two skills with the same name cannot sit side by side — one of them will be ignored.