agentleFS
Sign inStart free
AbdulrahmanAmerAbdulrahmanAmer/token-audit/skills/code-index/SKILL.md

code-index

Skill0 starsChanged 3 months agoMITView on GitHub

Generates a deterministic, greppable fact table — one line per fact — answering what you must know about a source file without opening it: what it is, its CLI, its internal imports, who breaks if you change an export, who runs it as a subprocess, who checks it, what it exports, and a file:line pointer to a load-bearing invariant. Use when asked who calls or depends on a file, what the blast radius of a change is, what a file does without reading it, or to set up or refresh a code index. Config-driven, derived never authored, and verified in CI with --check so it cannot drift. Saving is modest and measured — 1-3.4k tokens per fix, n=1, with a confound.

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. Chris-Carruthers/canon★ 4same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository3 other files it gives its agents.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Every fact kind has a test naming the defect it exists for.

    line 86

≈ “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

Code Index — a fact table you can trust without checking

One line per fact, answering "what must I know about this file without opening it?"

scripts/audit.mjs	IS	where did this session's tokens actually go?
scripts/audit.mjs	CLI	--file --json --list --no-paths --per-commit --project
scripts/audit.mjs	DEFINES	KINDS analyze classify encodeProject listTranscripts
scripts/audit.mjs	IMPORTEDBY	scripts/test/run-tests.mjs
scripts/audit.mjs	GUARD	scripts/test/run-tests.mjs
scripts/audit.mjs	WHY	scripts/audit.mjs:30 NO TOOL-RESULT CONTENT
kindanswers
ISwhat the file is, from its own header
CLIits command surface, from the argv dispatch
USESwhat it imports inside the repo
IMPORTEDBYblast radius, compile-time — who breaks if you change an export
SPAWNEDBYblast radius, runtime — who runs it as a subprocess
GUARDwho checks it
DEFINESexported names
WHYfile:line of a load-bearing invariant in its own comments

Say what it is worth, and do not oversell it

Measured saving in its home repo was 1–3.4k tokens per fix, n=1, with a real confound: the code was already in context for most of those fixes, so some of that saving is not attributable to the index. It is not the headline of this project — token-audit is. Say so if the user asks whether it is worth adopting. It is a modest, reliable saving on one shape of question — "who breaks if I change this" — otherwise answered by reading four files.

Running it

node ${CLAUDE_PLUGIN_ROOT}/scripts/code-index.mjs --root <repo>            # write it
node ${CLAUDE_PLUGIN_ROOT}/scripts/code-index.mjs --root <repo> --check    # CI: fail if stale
node ${CLAUDE_PLUGIN_ROOT}/scripts/code-index.mjs --root <repo> --stdout   # print, write nothing

Wire --check into CI in the same change that adds the index. Derived-but-unverified is strictly worse than nothing: the file drifts, and a stale fact table is a confident liar that people stop double-checking precisely because it has been right before.

Two properties, or it costs more than it saves

  1. Deterministic. Sorted lists, nothing from the clock, nothing from filesystem order, no generation date. It is meant to sit in a cached prompt prefix — one volatile byte at the top invalidates every token after it and turns the saving into a cost. If you are asked to add a timestamp or a file count to the header, refuse and explain this.
  2. Derived, never authored. Nobody edits it. If a fact is wrong, the generator or the config is wrong.

Config is where the value is

Roughly 10% of a generator like this binds to a repo's house conventions — the header format, the argv dispatch shape, where the guard register lives, how the codebase writes emphasis in comments — and that 10% is where all the value is. code-index.config.json at the repo root holds it; this plugin ships its own as the worked example, with every field commented.

When adopting it in a new repo, read the config first and fit it to that repo's conventions before generating anything. A default-config run produces a plausible table full of near misses, which is the worst possible first impression for a tool whose value is trust.

What it deliberately gets wrong in the safe direction

SPAWNEDBY under-reports. A spawn target is recorded only when a whitespace-free string literal reaches a spawn call directly or through one variable, and the target has a shebang. A path that reaches a spawn through a loop variable is not found. This is on purpose: a missing edge costs one grep, a wrong edge costs the trust that makes the table worth reading at all. Tell the user this rather than presenting the graph as complete.

If you change the generator

Every fact kind has a test naming the defect it exists for. Assert against a fresh build(), never against the committed artifact — a mutated generator never rewrites the file, so a suite that reads CODE-INDEX.txt stays green through any semantic breakage and only the staleness check fires. That trap made this tool's first test suite dead.

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