agentleFS
Sign inStart free
alilfrancesalilfrances/midas/skills/midas/SKILL.md

midas

Skill0 starsChanged 4 months agoMITView on GitHub

Midas overview + status. Trigger /midas for what plugin does, which gates active, how to disable (MIDAS_DISABLE=1).

The file

We split this skill into the rules it gives an agent

2 rules. About 0 are common, 2 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Verify gate: stop after edits with no check blocked once — verify or state why skipped.

    line 13
  2. Distinctive:

    Read both. Report: gates/nudges fired this session, top recurring lessons (n desc).

    line 30

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Midas

Scaffolding for mid-tier models. Active pieces:

  • Runtime coverage: Claude/Codex both run session_start, user_prompt, pre_tool, post_tool, post_tool_failure, and stop.
  • Hook entrypoints: Claude uses hooks/midas_hook.py; Codex uses hooks/codex_hook.py.
  • Session protocol injected at start (~150 tokens, once).
  • Edit gate: first edit with zero exploration denied once — grep first, retry.
  • Verify gate: stop after edits with no check blocked once — verify or state why skipped.
  • Thrash nudge: same command failing 2x triggers debug protocol pointer.
  • Large-read nudge: unbounded read >400 lines triggers grep-first pointer.
  • Bash router: bare cat/grep/find denied once per class; Claude uses Read/Grep/Glob, Codex uses rg plus bounded reads.
  • Freshness gate: API/version-looking failures or upgrade prompts nudge current-docs check.
  • Lessons: per-project thrash/fix pairs + midas-lesson notes inject top pitfalls from MIDAS_CONFIG_DIR/midas-data, else CLAUDE_CONFIG_DIR/midas-data, else ~/.claude/midas-data.
  • CLAUDE_PLUGIN_DATA is intentionally ignored for lessons because it is context-dependent and would split hook/CLI storage.
  • Scope: missing tools/access/network? state limit, offer nearest route, decline beats fake.

Each gate/nudge fires max once per session. Disable everything: MIDAS_DISABLE=1.

Playbooks (load on demand): midas:plan midas:explore midas:edit midas:debug midas:verify midas:ask midas:scope.

Self-review (on demand)

User asks how session went or /midas review:

  1. Locate lesson store with this precedence: MIDAS_CONFIG_DIR/midas-data, else CLAUDE_CONFIG_DIR/midas-data, else ~/.claude/midas-data. Lesson file: lessons-<sha1(realpath(cwd))[:12]>.json. Session state: $TMPDIR/midas-<session_id>.json.
  2. Read both. Report: gates/nudges fired this session, top recurring lessons (n desc).
  3. Suggest ONE concrete improvement max: CLAUDE.md rule, repo skill, or prompt habit — tied to highest-n lesson. No lesson n>=2 → say "no recurring failures, nothing to suggest". Zero cost until invoked — stats live in files, not context.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free