Receipt

Receipt — manage Change Receipts

Activation

Invoked via /receipt. Wraps lstack brain receipt CLI. The CLI is the source of truth. Never duplicate receipt storage logic here.

Supported forms

/receipt
/receipt help
/receipt start <task>
/receipt status
/receipt list
/receipt show <id>
/receipt finalize
/receipt finalize <summary>
/receipt abandon <reason>
/receipt explain
/receipt undo
/receipt record-test <command>
/receipt record-command <command>

Behavior by subcommand

/receipt or /receipt help

Run:

lstack brain receipt status

Show brief help listing the supported forms above. Do not create or finalize anything.

/receipt start

First check current state:

lstack brain receipt status

If no receipt is open, derive a short title (≤60 chars) and start one:

lstack brain receipt start --title "<short title>" --goal "<full task>"

If a receipt is already open, show it and use AskUserQuestion to ask:

• Continue the open receipt
• Finalize the open receipt first, then start new
• Abandon the open receipt first, then start new

Do not silently replace an open receipt.

/receipt status

lstack brain receipt status

/receipt list

lstack brain receipt list --limit 10

/receipt show

lstack brain receipt show <id>

<id> is a positional integer argument (not --id).

/receipt finalize or /receipt finalize

First:

lstack brain receipt status

If no receipt is open, report that clearly.

If a receipt is open:

1. Look for a test command in .claude/CLAUDE.md under Build & Test.
2. If a test command is found, ask the user whether to run it before finalizing.
3. If the user approves, run the test. Record the result honestly:

   lstack brain receipt record-test --command "<cmd>" --result pass
   # or
   lstack brain receipt record-test --command "<cmd>" --result fail
   

4. If no test command is known, either ask for one or ask whether to finalize without recorded tests.
5. Finalize:

   lstack brain receipt finalize --summary "<summary>"
   

   If no summary was supplied, generate one from the open receipt goal/title.

Never claim tests passed unless they actually passed. Never invent test results.

/receipt abandon

If reason is supplied:

lstack brain receipt abandon --reason "<reason>"

Note: --reason is required by the CLI; ask for it if missing.

If reason is missing, use AskUserQuestion to request one before proceeding.

/receipt explain

lstack brain receipt explain

/receipt undo

lstack brain receipt undo-hint

Print the suggested manual commands. Do not execute undo commands automatically.

/receipt record-test

Only run the command if it is clearly a test command or the user explicitly asked to run it. Record pass/fail honestly after actually running it:

lstack brain receipt record-test --command "<command>" --result pass
# or
lstack brain receipt record-test --command "<command>" --result fail

/receipt record-command

Only record commands that were actually run this session. Record result honestly.

lstack brain receipt record-command --command "<command>" --result pass
# or
lstack brain receipt record-command --command "<command>" --result fail

Constraints

• Never duplicate LBrain receipt storage logic in this skill.
• Never invoke Claude recursively or spawn subprocesses via AI CLI flags.
• Never run destructive or state-mutating git commands (reset, clean, restore, push, pull, commit).
• Never claim tests passed unless they actually passed.
• Never invent test results or receipt summaries.
• If the CLI returns an error, show it to the user and suggest what to run manually.
• The --reason flag is required by lstack brain receipt abandon — always supply it.
• lstack brain receipt show takes a positional integer, not --id.