Claude Skill

mix-compression

Reduce mix output noise (5-15% token savings) by installing rtk filters that compress mix test/credo/dialyzer/compile output before it reaches Claude. Use when long mix output floods context.

LLM Mart · 0 points · 0 views 0 listing impressions 0 install-command copies
Virus-scanned Reviewed automatically before listing.

Full trust report

Download oliver-kriska-claude-elixir-phoenix-plugins_elixir-phoenix_skills_mix-compression-9767a82.zip · 7 KB
oliver-kriska/claude-elixir-phoenix 560 44 forks MIT Updated 2d ago
Part of oliver-kriska/claude-elixir-phoenix — 93 skills

Install

skills CLI npx skills add https://github.com/oliver-kriska/claude-elixir-phoenix/tree/main/plugins/elixir-phoenix/skills/mix-compression
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install oliver-kriska-claude-elixir-phoenix@llmmart
Git git clone https://github.com/oliver-kriska/claude-elixir-phoenix.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole oliver-kriska/claude-elixir-phoenix collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

Mix Output Compression

Mix commands (mix test, mix credo, mix dialyzer, mix compile) emit verbose, repetitive output that consumes context fast. This skill installs rtk — a CLI proxy that filters tool output before it lands in the transcript.

The filters short-circuit happy paths to a single line (mix test: all pass) while preserving full failure blocks, compile errors, and stack traces. Net win: 5-15% per-session token reduction on mix-heavy workflows.

When to use

  • Long sessions — /phx:work or /phx:full hitting context limits from mix output
  • Debugging loops — /phx:investigate retrying mix compile/mix test repeatedly
  • Dialyzer-heavy projects — mix dialyzer output dominates the transcript

Iron Laws

  1. NEVER strip critical signals — compile errors (** (CompileError), == Compilation error in), test failures (FAILURES, 0 failures — preserved even on short-circuit), dialyzer warnings, and stack traces with file:line MUST pass through unchanged
  2. Verify after install — run rtk verify (or rtk verify --filter mix-test) to confirm the bundled test fixtures pass before declaring success
  3. Never overwrite existing .rtk/filters.toml — diff and merge instead

Workflow

Step 1: Detect rtk

which rtk && rtk --version

Read ${CLAUDE_SKILL_DIR}/references/install.md if rtk is missing — covers homebrew install + shell hook setup.

Step 2: Seed .rtk/filters.toml

Reference filters live at ${CLAUDE_SKILL_DIR}/references/rtk-filters.toml. Six production-tested filters covering:

  • mix-test — short-circuits all-pass, preserves failure blocks + compile errors
  • mix-credo — collapses clean runs, preserves violation blocks
  • mix-dialyzer — drops PLT progress, keeps warnings + summary
  • mix-deps-get — collapses unchanged package lists
  • mix-ecto-migrate — strips compile prefix, short-circuits "already up"
  • mix-compile — handles parallel worker prefixes (N>) and MIX_ENV

Run this if the project has no .rtk/filters.toml yet:

mkdir -p .rtk
cp "${CLAUDE_SKILL_DIR}/references/rtk-filters.toml" .rtk/filters.toml

Read both files if one already exists. Present a diff to the user. Merge only the filters they don't already have.

Step 3: Verify filters work

rtk verify                       # runs all embedded [[tests.*]] fixtures
rtk verify --filter mix-test     # one filter only

Check that all report "passed". Flag and stop if any fail — usually means the user has a custom rtk version with regex differences.

Step 4: Confirm shell hook

Run rtk init zsh (or rtk init bash) to install the transparent rewrite hook that turns mix X into rtk mix X. Re-running is safe (idempotent). Skip this step and mix calls run unfiltered.

Customization

Add custom regex patterns to strip_lines_matching for project-specific noise sources (e.g., third-party hex deps spamming stack traces). See the inline example in ${CLAUDE_SKILL_DIR}/references/rtk-filters.toml lines 57-59.

What this is NOT

  • Not a hook — Claude Code's PostToolUse hooks fire after the tool result is in the transcript and cannot shrink it. rtk works at the subprocess layer (the only layer where transcript-shortening is possible).
  • Not project-analysis — the bundled filter set is universal across Phoenix projects. No mix.exs inspection needed.
  • Not telemetry — rtk has telemetry off by default (enabled = false in config.toml). Filters run locally, no data leaves the machine.

References

  • ${CLAUDE_SKILL_DIR}/references/rtk-filters.toml — bundled filter set
  • ${CLAUDE_SKILL_DIR}/references/install.md — rtk install + shell hook setup
  • rtk on GitHub
Files (claude-elixir-phoenix)
  • references
    • install.md 3.3 KB 
      # Installing rtk

[rtk (Rust Token Killer)](https://github.com/rtk-ai/rtk) is a CLI proxy that
intercepts terminal commands, runs them, and filters output through TOML rules
before returning to the caller. For Claude Code, this means mix output gets
compressed *before* it reaches the transcript — the only architectural layer
where token savings are possible.

## Install

### macOS (homebrew)

```bash
brew install rtk-ai/tap/rtk
rtk --version    # verify (expect 0.39.0+)
```

### Other platforms

See the [rtk README](https://github.com/rtk-ai/rtk#install) for Linux and
manual install options.

## Shell hook (required)

rtk only intercepts commands when its shell hook is installed. The hook
rewrites `mix X` to `rtk mix X` transparently — no manual prefixing.

```bash
rtk init zsh  # or `rtk init bash`
```

This adds a hook to `~/.zshrc` (or `~/.bashrc`). Reload your shell:

```bash
exec $SHELL -l
```

Verify the hook works:

```bash
mix --help    # should run normally — rtk passes through commands without filters
```

If you see `rtk: command not found` or `mix` runs raw without filtering, the
hook didn't install. Check `~/.zshrc` for an `rtk init` block.

## Verify filters

After dropping filters into `.rtk/filters.toml`:

```bash
rtk verify                              # run all embedded [[tests.*]] fixtures
rtk verify --filter mix-test            # run fixtures for one filter only
rtk verify --filter mix-ash-codegen
```

Each filter should report all fixtures as passed. If any fail, the regex
patterns may not match your rtk version's regex engine — file an issue on
the rtk repo.

## Measure the win

```bash
rtk gain                  # token savings summary
rtk gain --history        # per-command breakdown
rtk cc-economics          # spending vs savings analysis
```

Run after a week of normal usage to see actual reduction. Expected ranges
for Phoenix projects:

| Command | Typical compression | Big win cases |
|---------|-------------------:|---------------|
| `mix test` (passing) | 90%+ | All-green runs collapse to one line |
| `mix credo` (clean) | 95%+ | Clean runs collapse to one line |
| `mix dialyzer` (clean) | 95%+ | PLT progress disappears |
| `mix compile` (no warnings) | 70-90% | Worker-prefix noise stripped |
| `mix test` (failures) | 10-30% | Failure blocks preserved verbatim |

## Telemetry

rtk has telemetry **disabled by default**. Verify in
`~/Library/Application Support/rtk/config.toml`:

```toml
[telemetry]
enabled = false
consent_given = false
```

Filters run locally. No command output, file paths, or filter content leaves
your machine.

## Disable / uninstall

To disable temporarily without uninstalling:

```bash
# Remove the rtk init line from ~/.zshrc, then:
exec $SHELL -l
```

To uninstall:

```bash
brew uninstall rtk
rm -rf ~/Library/Application\ Support/rtk
rm -rf .rtk    # in any project that has one
```

## Troubleshooting

**Filter not applying**: rtk only filters when the command matches a filter's
`match_command` regex. Check `rtk config show-filters` to list registered
filters in the current directory.

**Test fixtures fail after copying**: rtk uses Rust's regex crate which does
NOT support backreferences (`\1`, `\2`). If you wrote your own patterns, avoid
them.

**Want raw output for one command**: prefix with `command`:

```bash
command mix test    # bypasses rtk
```
    • rtk-filters.toml 15.6 KB 
      schema_version = 1

# RTK TOML filters for Elixir mix commands.
#
# Drop into `.rtk/filters.toml` at the project root. Filters apply when mix
# commands run via `rtk mix ...` (rtk's shell hook rewrites `mix X` -> `rtk mix X`
# transparently — see references/install.md).
#
# Design notes:
# - Every mix-* filter includes shared compile-strip patterns because parallel
#   deps.compile often runs as a side effect, emitting "Compiling N files" and
#   "Generated X app" noise with worker-prefix form `N> `.
# - filter_stderr = true because mix emits warnings to stderr; we want them
#   filtered through the same pipeline, not leaking raw.
# - match_output short-circuits the "happy path" to a one-liner before any
#   other processing — biggest wins for `mix test` (passing runs are the norm).
# - Rust regex crate (RTK's engine) does NOT support backreferences (\1, \2).
#   Any pattern using them will silently fail.
# - Critical-preserve: compile errors, test failures, dialyzer warnings, and
#   stack traces with file:line are NEVER stripped — only progress noise is.

# ============================================================================
# mix test — compact pass/fail, preserve failure blocks, stacktraces, compile errors
# ============================================================================
[filters.mix-test]
description = "Compact mix test — short-circuit all-green, preserve failures and compile errors"
match_command = "^(?:[A-Z_][A-Z0-9_]*=\\S+\\s+)*mix\\s+test(\\s|$)"
strip_ansi = true
filter_stderr = true

match_output = [
  { pattern = "\\b0 failures\\b", message = "mix test: all pass" },
]

replace = [
  { pattern = "^[.*sF]{3,}", replacement = "" },
]

strip_lines_matching = [
  "^$",
  "^\\s*$",
  "^Excluding tags:",
  "^Including tags:",
  "^Randomized with seed",
  "^OK$",
  "^Compiling \\d+ file",
  "^Generated \\w+ app$",
  "^\\d+> Compiling \\d+ file",
  "^\\d+> Generated \\w+ app",
  "^\\d+> ==> \\w+",
  "^\\d+> ===> (Compiling|Analyzing)",
  "^==> \\w+$",
  "^build ended with \\d+ errors$",
  "^\\d{2}:\\d{2}:\\d{2}\\.\\d+ ",
  "^\\d{2}:\\d{2}:\\d{2}\\.\\d+ app_version=",
  "^\\d{2}:\\d{2}:\\d{2}\\.\\d+ line=\\d+",
  # Stack-trace frame strippers — drop frames from libraries you don't care about.
  # Add your own deps here: e.g., "^\\s+\\(my_dep \\d" strips lines like "    (my_dep 1.2.3) ..."
  "^\\s+\\(phoenix_live_view \\d",
  "^\\s+\\(elixir \\d",
  "^\\s+\\(ecto \\d",
  "^warning: Range\\.new",
  "^warning: redefining module ",
  "^Query: INSERT INTO",
  "^Query: UPDATE",
  "^Query: SELECT",
  "^Params: ",
  "^Reason: %Postgrex",
  "^Function: #Function<",
  "^    Args: \\[\\]",
]

tail_lines = 200
on_empty = "mix test: ok"

[[tests.mix-test]]
name = "all pass short-circuits"
input = """
Running ExUnit with seed: 986129, max_cases: 28
Excluding tags: [:ci_only]

.
Finished in 1.1 seconds (0.00s async, 1.1s sync)
1 test, 0 failures

OK
"""
expected = "mix test: all pass"

[[tests.mix-test]]
name = "compile error preserved"
input = """
Running ExUnit with seed: 755757, max_cases: 28
Excluding tags: [:ci_only]

    error: undefined variable ^b_item
    │
 20 │       assert [{^a_item, ^b_item}] = [Enum.at(a, 0)]
    │                          ^
    │
    └─ test/validation_match_test.exs:20:26: ValidationMatchTest

== Compilation error in file test/validation_match_test.exs ==
** (CompileError) test/validation_match_test.exs: cannot compile module
"""
expected = """Running ExUnit with seed: 755757, max_cases: 28
    error: undefined variable ^b_item
    │
 20 │       assert [{^a_item, ^b_item}] = [Enum.at(a, 0)]
    │                          ^
    │
    └─ test/validation_match_test.exs:20:26: ValidationMatchTest
== Compilation error in file test/validation_match_test.exs ==
** (CompileError) test/validation_match_test.exs: cannot compile module"""

# ============================================================================
# mix credo — drop redefining-module warnings, progress lines, keep violations
# ============================================================================
[filters.mix-credo]
description = "Compact mix credo — drop custom-check redefine warnings and progress"
match_command = "^(?:[A-Z_][A-Z0-9_]*=\\S+\\s+)*mix\\s+credo(\\s|$)"
strip_ansi = true
filter_stderr = true

match_output = [
  { pattern = "found no issues\\.", message = "mix credo: clean" },
]

strip_lines_matching = [
  "^$",
  "^\\s*$",
  "^OK$",
  "^Compiling \\d+ file",
  "^Generated \\w+ app$",
  "^\\d+> Compiling",
  "^\\d+> Generated",
  "^\\s*warning: redefining module ",
  "^\\s+│$",
  "^\\s+~+$",
  "^  \\d+ │ defmodule ",
  "^\\s+└─ lib/mix/credo/",
  "^Checking \\d+ source files",
  "^Please report incorrect results:",
  "^Analysis took \\d",
  "^Use `mix credo explain`",
]

on_empty = "mix credo: clean"

[[tests.mix-credo]]
name = "clean run short-circuits"
input = """
    warning: redefining module MyApp.Test.Credo.WarnApplicationConfig
    │
  1 │ defmodule MyApp.Test.Credo.WarnApplicationConfig do
    │ ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
    │
    └─ lib/mix/credo/warn_application_config.ex:1: MyApp.Test.Credo.WarnApplicationConfig (module)

Checking 2579 source files (this might take a while) ...
Analysis took 13.4 seconds
29543 mods/funs, found no issues.

Use `mix credo explain` to explain issues.
OK
"""
expected = "mix credo: clean"

[[tests.mix-credo]]
name = "violation block preserved"
input = """
Compiling 2 files (.ex)
Generated my_app app

  Code Readability
┃
┃ [R] ↗ Variable names should be written in snake_case.
┃       config/runtime.exs:16:1 #()
┃ [R] ↘ Line is too long (max is 120, was 158).
┃       config/runtime.exs:85:121 #()

Please report incorrect results: https://github.com/rrrene/credo/issues
Analysis took 0.3 seconds
0 mods/funs, found 2 code readability issues.

Use `mix credo explain` to explain issues.
"""
expected = """  Code Readability
┃
┃ [R] ↗ Variable names should be written in snake_case.
┃       config/runtime.exs:16:1 #()
┃ [R] ↘ Line is too long (max is 120, was 158).
┃       config/runtime.exs:85:121 #()
0 mods/funs, found 2 code readability issues."""

# ============================================================================
# mix dialyzer — drop PLT progress and config dumps, keep warnings + summary
# ============================================================================
[filters.mix-dialyzer]
description = "Compact mix dialyzer — drop PLT progress, keep warnings and summary"
match_command = "^(?:[A-Z_][A-Z0-9_]*=\\S+\\s+)*mix\\s+dialyzer(\\s|$)"
strip_ansi = true
filter_stderr = true

match_output = [
  { pattern = "done \\(passed successfully\\)", message = "mix dialyzer: clean" },
]

strip_lines_matching = [
  "^$",
  "^\\s*$",
  "^OK$",
  "^Compiling \\d+ file",
  "^Generated \\w+ app$",
  "^Finding suitable PLTs",
  "^Checking PLT\\.\\.\\.",
  "^Looking up modules in ",
  "^Finding applications for ",
  "^Finding modules for ",
  "^Checking \\d+ modules in ",
  "^Adding \\d+ modules to ",
  "^Removing \\d+ modules from ",
  "^done in \\d+m?\\d+",
  "^ignore_warnings: ",
  "^Starting Dialyzer",
  "^\\[$",
  "^  check_plt: ",
  "^  init_plt: ",
  "^  files: ",
  "^   ~c\"",
  "^   \\.\\.\\.\\],",
  "^  \\.\\.\\.",
  "^\\]$",
  "^\\[:[a-z_]+,",
  "^_{40,}$",
  "^Halting VM with exit status \\d",
]

on_empty = "mix dialyzer: clean"

[[tests.mix-dialyzer]]
name = "clean run short-circuits"
input = """
Generated my_app app
Finding suitable PLTs
Checking PLT...
[:abnf_parsec, :acceptor_pool, :algoliax, :apollo_io]
Looking up modules in dialyxir_erlang-_elixir-_deps-dev.plt
Checking 9618 modules in dialyxir_erlang-_elixir-_deps-dev.plt
Adding 223 modules to dialyxir_erlang-_elixir-_deps-dev.plt
done in 0m51.82s
ignore_warnings: .dialyzer_ignore_warnings.exs

Starting Dialyzer
[
  check_plt: false,
  init_plt: ~c"priv/plts/dev/dialyxir_erlang.plt",
  files: [~c"/Users/.../my_app.beam",
   ...],
  ...
]
Total errors: 156, Skipped: 156, Unnecessary Skips: 0
done in 0m27.35s
done (passed successfully)
OK
"""
expected = "mix dialyzer: clean"

[[tests.mix-dialyzer]]
name = "warnings preserved dividers stripped"
input = """
Total errors: 156, Skipped: 154, Unnecessary Skips: 0
done in 0m40.32s
________________________________________________________________________________
lib/my_app/reports/excel_formatting.ex:407:14:unknown_function
Function Elixlsx.Sheet.set_col_width/3 does not exist.
________________________________________________________________________________
lib/my_app/reports/excel_formatting.ex:418:13:unknown_function
Function Elixlsx.Sheet.set_col_width/3 does not exist.
________________________________________________________________________________
done (warnings were emitted)
Halting VM with exit status 2
"""
expected = """Total errors: 156, Skipped: 154, Unnecessary Skips: 0
lib/my_app/reports/excel_formatting.ex:407:14:unknown_function
Function Elixlsx.Sheet.set_col_width/3 does not exist.
lib/my_app/reports/excel_formatting.ex:418:13:unknown_function
Function Elixlsx.Sheet.set_col_width/3 does not exist.
done (warnings were emitted)"""

# ============================================================================
# mix deps.get — collapse "Unchanged" package list
# ============================================================================
[filters.mix-deps-get]
description = "Compact mix deps.get — collapse unchanged package lists"
match_command = "^(?:[A-Z_][A-Z0-9_]*=\\S+\\s+)*mix\\s+deps\\.get(\\s|$)"
strip_ansi = true
filter_stderr = true

match_output = [
  { pattern = "All dependencies (have been fetched|are up to date)", message = "mix deps.get: unchanged" },
]

strip_lines_matching = [
  "^$",
  "^\\s*$",
  "^OK$",
  "^Resolving Hex dependencies",
  "^Resolution completed in ",
  "^Unchanged:$",
  "^  [a-z_][a-z0-9_]* \\d+\\.\\d+",
  "^Compiling \\d+ file",
  "^Generated \\w+ app$",
]

on_empty = "mix deps.get: unchanged"

[[tests.mix-deps-get]]
name = "unchanged collapses"
input = """
Resolving Hex dependencies...
Resolution completed in 1.638s
Unchanged:
  abnf_parsec 2.1.0
  acceptor_pool 1.0.1
  algoliax 0.10.0
  credo 1.7.18
All dependencies have been fetched
OK
"""
expected = "mix deps.get: unchanged"

# ============================================================================
# mix ecto.migrate / rollback / gen.migration — strip compile prefix
# ============================================================================
[filters.mix-ecto-migrate]
description = "Compact mix ecto.migrate — strip compile prefix"
match_command = "^(?:[A-Z_][A-Z0-9_]*=\\S+\\s+)*mix\\s+ecto\\.(migrate|rollback|gen\\.migration)(\\s|$)"
strip_ansi = true
filter_stderr = true

match_output = [
  { pattern = "\\[info\\] Migrations already up", message = "mix ecto.migrate: up to date" },
]

strip_lines_matching = [
  "^$",
  "^\\s*$",
  "^OK$",
  "^Compiling \\d+ file",
  "^Generated \\w+ app$",
  "^\\d+> Compiling",
  "^\\d+> Generated",
]

on_empty = "mix ecto.migrate: ok"

[[tests.mix-ecto-migrate]]
name = "already up short-circuits"
input = """
Compiling 2 files (.ex)
Generated my_app app

06:54:27.843 [info] Migrations already up
"""
expected = "mix ecto.migrate: up to date"

# ============================================================================
# mix compile (extended) — handles parallel worker prefix `N> ` and MIX_ENV
# Shadows the built-in mix-compile filter.
# ============================================================================
[filters.mix-compile]
description = "Compact mix compile — handles parallel workers and MIX_ENV prefixes"
match_command = "^(?:[A-Z_][A-Z0-9_]*=\\S+\\s+)*mix\\s+compile(\\s|$)"
strip_ansi = true
filter_stderr = true

replace = [
  { pattern = "^\\d+> ", replacement = "" },
]

strip_lines_matching = [
  "^$",
  "^\\s*$",
  "^OK$",
  "^mix deps\\.compile running across \\d+ OS processes",
  "^Compiling \\d+ file",
  "^Generated \\w+ app$",
  "^==> \\w+$",
  "^===> (Compiling|Analyzing) ",
  "^\\[debug\\] Copying NIF ",
  "^AppSignal extension installation",
]

on_empty = "mix compile: ok"

[[tests.mix-compile]]
name = "parallel compile with worker prefix strips cleanly"
input = """
mix deps.compile running across 7 OS processes
2> ==> peri
2> Compiling 7 files (.ex)
3> ==> fine
3> Compiling 1 file (.ex)
3> Generated fine app
2> Generated peri app
Compiling 1821 files (.ex)
Generated my_app app
OK
"""
expected = "mix compile: ok"

[[tests.mix-compile]]
name = "preserves warning with file:line"
input = """
2> ==> peri
2> Compiling 7 files (.ex)
3>     warning: variable "assigns" is unused
3>     │
3> 183 │     defp prepare_icon_name(assigns) do
3>     │                            ~~~~~~~
3>     │
3>     └─ lib/icon.ex:183:28: Iconify.Icon.prepare_icon_name/1
2> Generated peri app
Compiling 1821 files (.ex)
Generated my_app app
"""
expected = """    warning: variable "assigns" is unused
    │
183 │     defp prepare_icon_name(assigns) do
    │                            ~~~~~~~
    │
    └─ lib/icon.ex:183:28: Iconify.Icon.prepare_icon_name/1"""

# ============================================================================
# mix ash.codegen — strip compile prefix, preserve snapshot/migration lists
# ============================================================================
[filters.mix-ash-codegen]
description = "Compact mix ash.codegen — strip compile prefix, preserve snapshot and migration file lists"
match_command = "^(?:[A-Z_][A-Z0-9_]*=\\S+\\s+)*mix\\s+ash\\.codegen(\\s|$)"
strip_ansi = true
filter_stderr = true

match_output = [
  { pattern = "No changes", message = "mix ash.codegen: no changes" },
]

strip_lines_matching = [
  "^$",
  "^\\s*$",
  "^OK$",
  "^Compiling \\d+ file",
  "^Generated \\w+ app$",
  "^\\d+> Compiling",
  "^\\d+> Generated",
]

on_empty = "mix ash.codegen: ok"

[[tests.mix-ash-codegen]]
name = "no changes short-circuits"
input = """
Compiling 5 files (.ex)
Generated my_app app

No changes detected.
OK
"""
expected = "mix ash.codegen: no changes"

[[tests.mix-ash-codegen]]
name = "snapshot and migration list preserved"
input = """
Compiling 5 files (.ex)
Generated my_app app

Snapshots Updated:
  * priv/resource_snapshots/repo/accounts/user_v2.json

Migrations Generated:
  * priv/repo/migrations/20260511120000_migrate_resources2.exs
"""
expected = """Snapshots Updated:
  * priv/resource_snapshots/repo/accounts/user_v2.json
Migrations Generated:
  * priv/repo/migrations/20260511120000_migrate_resources2.exs"""

# ============================================================================
# mix ash.migrate — strip compile prefix, preserve migration steps
# ============================================================================
[filters.mix-ash-migrate]
description = "Compact mix ash.migrate — strip compile prefix, preserve migration steps"
match_command = "^(?:[A-Z_][A-Z0-9_]*=\\S+\\s+)*mix\\s+ash\\.(migrate|rollback)(\\s|$)"
strip_ansi = true
filter_stderr = true

match_output = [
  { pattern = "\\[info\\] Migrations already up", message = "mix ash.migrate: up to date" },
]

strip_lines_matching = [
  "^$",
  "^\\s*$",
  "^OK$",
  "^Compiling \\d+ file",
  "^Generated \\w+ app$",
  "^\\d+> Compiling",
  "^\\d+> Generated",
]

on_empty = "mix ash.migrate: ok"

[[tests.mix-ash-migrate]]
name = "already up short-circuits"
input = """
Compiling 2 files (.ex)
Generated my_app app

06:54:27.843 [info] Migrations already up
"""
expected = "mix ash.migrate: up to date"

[[tests.mix-ash-migrate]]
name = "migration steps preserved"
input = """
Compiling 2 files (.ex)
Generated my_app app

07:00:00.000 [info] == Running 20260511120000 MyApp.Repo.Migrations.MigrateResources2.up/0 forward
07:00:00.000 [info] == Migrated 20260511120000 in 0.0s
"""
expected = """07:00:00.000 [info] == Running 20260511120000 MyApp.Repo.Migrations.MigrateResources2.up/0 forward
07:00:00.000 [info] == Migrated 20260511120000 in 0.0s"""
  • SKILL.md 4 KB 
    ---
name: mix-compression
description: Reduce mix output noise (5-15% token savings) by installing rtk filters that compress mix test/credo/dialyzer/compile output before it reaches Claude. Use when long mix output floods context.
effort: low
---

# Mix Output Compression

Mix commands (`mix test`, `mix credo`, `mix dialyzer`, `mix compile`) emit
verbose, repetitive output that consumes context fast. This skill installs
[rtk](https://github.com/rtk-ai/rtk) — a CLI proxy that filters tool output
**before it lands in the transcript**.

The filters short-circuit happy paths to a single line (`mix test: all pass`)
while preserving full failure blocks, compile errors, and stack traces. Net win:
5-15% per-session token reduction on mix-heavy workflows.

## When to use

- **Long sessions** — `/phx:work` or `/phx:full` hitting context limits from mix output
- **Debugging loops** — `/phx:investigate` retrying `mix compile`/`mix test` repeatedly
- **Dialyzer-heavy projects** — `mix dialyzer` output dominates the transcript

## Iron Laws

1. **NEVER strip critical signals** — compile errors (`** (CompileError)`,
   `== Compilation error in`), test failures (`FAILURES`, `0 failures` — preserved
   even on short-circuit), dialyzer warnings, and stack traces with `file:line` MUST
   pass through unchanged
2. **Verify after install** — run `rtk verify` (or `rtk verify --filter mix-test`) to
   confirm the bundled test fixtures pass before declaring success
3. **Never overwrite existing `.rtk/filters.toml`** — diff and merge instead

## Workflow

### Step 1: Detect rtk

```bash
which rtk && rtk --version
```

Read `${CLAUDE_SKILL_DIR}/references/install.md` if rtk is missing — covers
homebrew install + shell hook setup.

### Step 2: Seed `.rtk/filters.toml`

Reference filters live at `${CLAUDE_SKILL_DIR}/references/rtk-filters.toml`. Six
production-tested filters covering:

- **`mix-test`** — short-circuits all-pass, preserves failure blocks + compile errors
- **`mix-credo`** — collapses clean runs, preserves violation blocks
- **`mix-dialyzer`** — drops PLT progress, keeps warnings + summary
- **`mix-deps-get`** — collapses unchanged package lists
- **`mix-ecto-migrate`** — strips compile prefix, short-circuits "already up"
- **`mix-compile`** — handles parallel worker prefixes (`N>`) and MIX_ENV

Run this if the project has no `.rtk/filters.toml` yet:

```bash
mkdir -p .rtk
cp "${CLAUDE_SKILL_DIR}/references/rtk-filters.toml" .rtk/filters.toml
```

Read both files if one already exists. Present a diff to the user. Merge only
the filters they don't already have.

### Step 3: Verify filters work

```bash
rtk verify                       # runs all embedded [[tests.*]] fixtures
rtk verify --filter mix-test     # one filter only
```

Check that all report "passed". Flag and stop if any fail — usually means the
user has a custom rtk version with regex differences.

### Step 4: Confirm shell hook

Run `rtk init zsh` (or `rtk init bash`) to install the transparent rewrite hook
that turns `mix X` into `rtk mix X`. Re-running is safe (idempotent). Skip this
step and `mix` calls run unfiltered.

## Customization

Add custom regex patterns to `strip_lines_matching` for project-specific noise
sources (e.g., third-party hex deps spamming stack traces). See the inline
example in `${CLAUDE_SKILL_DIR}/references/rtk-filters.toml` lines 57-59.

## What this is NOT

- **Not a hook** — Claude Code's `PostToolUse` hooks fire after the tool result
  is in the transcript and cannot shrink it. rtk works at the subprocess layer
  (the only layer where transcript-shortening is possible).
- **Not project-analysis** — the bundled filter set is universal across Phoenix
  projects. No `mix.exs` inspection needed.
- **Not telemetry** — rtk has telemetry off by default (`enabled = false` in
  `config.toml`). Filters run locally, no data leaves the machine.

## References

- `${CLAUDE_SKILL_DIR}/references/rtk-filters.toml` — bundled filter set
- `${CLAUDE_SKILL_DIR}/references/install.md` — rtk install + shell hook setup
- [rtk on GitHub](https://github.com/rtk-ai/rtk)

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.

Related

Claude Prompt
Tighten-my-prose editing prompt
Claude Skill
Adversarial code-review skill for Claude
Claude Workflow
Deep-research workflow (fan-out + verify)
Claude Skill
Conventional-commit writer for Claude