credo

name	credo
description	Run Credo static analysis on Elixir projects. Dispatches subcommands via `/credo [subcommand]`. Use when the user says '/credo check', '/credo fix', '/credo help', 'run credo', 'credo issues', or wants to check or fix Credo warnings in an Elixir project.

Credo

Run Credo static analysis checks and fix issues in Elixir projects. Expects Credo to be a dependency in the project's mix.exs.

Subcommands

Subcommand	Purpose
check	Run mix credo --strict and report issues
fix	Run Credo, then fix all reported issues

/credo help

Display a list of all available subcommands. Output the following exactly:

/credo subcommands:

  check   — Run mix credo --strict and report issues
  fix     — Run Credo, then fix all reported issues
  help    — Show this help message

If /credo is invoked without a subcommand, default to check.

Dispatch

1. Parse the subcommand from the user's invocation. Examples:
   • /credo → default to check
   • /credo check → subcommand check
   • /credo fix → subcommand fix
   • /credo help → show help
2. If the subcommand is unknown, list available subcommands and stop.
3. Follow the matching workflow below.

/credo check

Run Credo analysis and report results.

1. Run mix credo --strict in the project root.
2. If Credo exits cleanly with no issues, report success and stop.
3. If issues are found, present a concise summary:
   • Total issue count grouped by category (Consistency, Design, Readability, Refactor, Warning)
   • List each issue with file, line, and message
4. Do NOT fix anything — just report.

/credo fix

Run Credo and fix all reported issues.

1. Check for existing Credo output — if /credo check was already run in this conversation and the results are still available:
   • Show the user the issue count from the previous run.
   • Ask the user: "Credo check was already run with N issues found. Do you want to re-run the check or proceed with fixing the existing issues?"
   • If the user chooses to proceed, skip to step 3 using the existing results.
   • If the user chooses to re-run, continue to step 2.
2. Run mix credo --strict --format json in the project root. If Credo exits cleanly with no issues, report success and stop.
3. If issues are found:
   • Parse the JSON output to get the list of issues with file paths, line numbers, check names, and messages.
   • Group issues by file.
   • For each file, read the file and fix all Credo issues in that file:
     • Consistency issues (e.g., SpaceAroundOperators, SpaceInParentheses): apply the consistent style.
     • Readability issues (e.g., AliasOrder): reorder aliases alphabetically. For ModuleDoc: add @moduledoc false if the module is internal/private, otherwise add a meaningful @moduledoc. For MaxLineLength: break long lines.
     • Refactor issues (e.g., Nesting, CyclomaticComplexity): refactor to reduce complexity. For WithClauses: simplify with blocks. For Nesting: extract helper functions.
     • Warning issues (e.g., IoInspect, IExPry): remove debug statements. For UnusedEnumOperation and similar: ensure return values are used or remove dead code.
     • Design issues (e.g., AliasUsage): add missing aliases. For TagTODO/TagFIXME: ask the user whether to resolve or suppress.
     • ExSlop issues (AI slop detection): fix the specific anti-pattern identified by the check message.
   • After fixing all issues in a file, save the file.
4. After all files are fixed, run mix credo --strict again to verify.
5. If issues remain, report them and attempt a second pass on the remaining issues.
6. Report final results: how many issues were fixed, how many remain (if any).

Notes

• Always use --strict flag to catch all issues including low-priority ones.
• For --format json, if the project's Credo version doesn't support it, fall back to parsing the standard text output.
• When fixing, preserve the existing code style and intent — only change what Credo flags.
• If a fix would change behavior (not just style), ask the user before applying.
• ExSlop checks (AI slop detection) are custom checks from the ex_slop dependency — treat them like standard Credo checks.

$ARGUMENTS