Decision Ledger Claude Code mod: a Claude Code decision log for a project

Decision Ledger is a community mod by YeonwooSung (YeonwooSung), one plugin in the my-claude-code-mods collection. It is a small Claude Code decision log. When you settle a design question during a session, you write down the choice, the reason and the option you turned down, and attach it to a file or symbol such as src/cache.ts#lookup. Later you can list everything, filter by file, or remove an entry.

It adds one slash command, /decision:

• /decision add <file#symbol> | <choice> | <rationale> | <rejected alternative> records a decision. The rejected alternative is optional.
• /decision list prints every decision, and /decision list src/cache.ts prints only those whose reference contains that text.
• /decision remove <id> deletes one entry.
• /decision reset clears this mod's records.
• /decision help prints the syntax.

The reply appears as command text in the transcript, so it works in the terminal, in Desktop and in headless sessions. Each entry is printed as an id and reference, then Choice, Why and Rejected lines. Nothing is recorded unless you type it.

Use any of the methods in the install block; they install straight from the author's own marketplace, my-claude-code-mods. Reload plugins afterwards (or restart Claude Code) so /decision is registered. To try it without installing, point --plugin-dir at the decision-ledger folder of a checkout:

claude --plugin-dir ./decision-ledger

The plugin is one hooks module, hooks/register.ts, plus a data class in hooks/core/feature.ts and shared helpers in hooks/core/runtime.ts and hooks/core/redact.ts. On session.start it registers the command. On command.run it loads the saved state from $.store, hands the arguments to the Feature class, and writes the state back only if it changed. The add branch, from feature.ts:

if (verb === 'add') {
  const [reference, choice, rationale, rejected = ''] = entries(text)
  if (!reference || !choice || !rationale) return 'Usage: ' + HELP
  this.data.decisions = bounded([...this.data.decisions, { id: this.data.nextId++, reference: safeText(reference, 500), choice: safeText(choice, 1000), rationale: safeText(rationale, 2000), rejected: safeText(rejected, 1000) }])
}

entries splits on |, safeText strips terminal escape sequences and masks obvious secrets, and bounded keeps the last 200. Commands run one at a time through a queue, and the key includes the working directory and session id, so two folders never share a ledger. If the save fails, the command reports an error and the stored state is left as it was.

There are no options in /config. The limits are constants in the source: 200 entries (bounded in runtime.ts), per-field length caps of 500, 1000, 2000 and 1000 characters (feature.ts), and a 256 KiB cap on saved state (register.ts). Edit them in a local checkout and load it with --plugin-dir if you want different ones.

We read every file the plugin can execute before running it: .claude-plugin/plugin.json, hooks/hooks.json, hooks/register.ts, hooks/core/feature.ts, hooks/core/runtime.ts, hooks/core/redact.ts, and the two test files. There is no package.json inside the plugin, no install script, no bin/, no bundled or minified code and no binary. Findings:

• Network: none. We searched for fetch, http and child_process and found nothing. No telemetry.
• Processes: register.ts builds a helper object with $.process.run, but the only command, Feature.command, never calls it. The helpers that would run git (runtime.ts) and node -e (register.ts:24) are unused by this mod. They are shared code from the author's other plugins.
• Files: no direct writes. $.fs.read, $.fs.stat and $.fs.list are wired into the same unused helper object. State goes through $.store, the plugin's own store; nothing is written to ~/.claude, settings, shell files or LaunchAgents.
• Daemons: none.
• Secrets: it reads no environment variables or credential files. Text you enter is passed through redactSecrets, which masks things like --password, bearer tokens, KEY= assignments and URL passwords. The README calls this best effort, so do not type secrets into a rationale.
• Prompt influence: none. It injects nothing into prompts and returns only the command's own text.
• Tool hook: tool.call is registered, but because Feature has no observe method (register.ts:9) it only calls next(e) and returns. It never blocks or rewrites a tool call.
• Requirements: none beyond Claude Code with function hooks.
• License: Apache-2.0, with a LICENSE file at the repository root.

claude plugin validate passed. It listed two notes that the command.run and tool.call hooks have no .catch, and one non-blocking warning from its directory check that the plugin has no icon. The author's 10 tests (2 files) passed with claude plugin test on Claude Code 2.1.307.

Checked with Claude Code 2.1.307 on macOS (validate and tests). The author tested on 2.1.302 and says the live terminal and Desktop rendering was not checked. If /decision does not respond:

• Run /reload-plugins or restart Claude Code after installing.
• Run /decision help to see the syntax. A missing choice or rationale returns the usage line instead of saving.
• Use | only as the separator; the add command cannot hold a pipe inside a field.
• "Session changed; operation not applied" means the folder or session id changed while the command was running. Run it again.
• An empty ledger after a restart usually means a different folder or session id, because entries are keyed on both.

To check that it works, run /decision add a.ts#f | choice | reason, then /decision list.