receipts
Counts every time a coding agent said work was done, fixed, tested, committed or pushed, and checks each claim against the same session's own commands, marking it Confirmed, Contradicted or Unverified. Runs the open-source Receipts collector locally over Claude Code and Codex session logs and prints the counts as a 40-column card the user can share. Reads local files only and sends nothing anywhere. Use when the user asks how often their agent's "done" was true, wants to audit agent completion claims, or asks for their receipts card.
- License 许可
- MIT
- 最近更新
- 2026-11-14
- Stars
- ★ 1.6k
- Forks
- 491
01 前言 Front Matter
- name
- receipts
- license
- MIT
- metadata
- { "author": "Nikxius, Inc.", "source": "https://github.com/Kloudedge-apex/receipts-collector", "version": "0.1.0", "homepage": "https://receipts.nueton.ai" }
- description
- Counts every time a coding agent said work was done, fixed, tested, committed or pushed, and checks each claim against the same session's own commands, marking it Confirmed, Contradicted or Unverified. Runs the open-source Receipts collector locally over Claude Code and Codex session logs and prints the counts as a 40-column card the user can share. Reads local files only and sends nothing anywhere. Use when the user asks how often their agent's "done" was true, wants to audit agent completion claims, or asks for their receipts card.
- allowed-tools
- Bash(python3 ${CLAUDE_SKILL_DIR}/scripts/receipts_collect.py *)
- compatibility
- Requires Python 3.9 or newer and local Claude Code (~/.claude/projects) or Codex (~/.codex/sessions) logs. Runs offline; needs no network access, account or API key.
02 资源文件 1
- 脚本scripts/receipts_collect.py60.5 KB
03 SKILL.md 正文
Receipts
Your agent said "done". This counts how many of those times its own logs back it up.
The collector is scripts/receipts_collect.py next to this file: one Python file, standard library only, MIT licensed. In Claude Code, ${CLAUDE_SKILL_DIR} expands to this skill's folder. If you see the variable unexpanded (Codex and other agents), use the absolute path of the scripts/ folder beside this SKILL.md instead.
Steps
-
Tell the user, in one sentence, what is about to happen: the collector reads session logs under
~/.claude/projectsand~/.codex/sessions, writes a hashed export to~/.nueton-receipts/export.json(and a private salt file at~/.nueton-receipts/salton first run), and sends nothing anywhere. -
Run the collector:
python3 ${CLAUDE_SKILL_DIR}/scripts/receipts_collect.py --summary --cardThe card prints on stdout. The one-line summary and the diagnostic counts print on stderr.
Add flags only when the user asks for them:
--since YYYY-MM-DDand--until YYYY-MM-DDfor a date range (default: the 30 days ending at the newest log record).--claude-dir PATHor--codex-dir PATHwhen the logs live somewhere else.--repo-checkto let commit and push claims use read-only local Git history.--no-sentencesto keep claim text out of the export entirely.--card-out PATHto also save the card as a text file.
If
python3is missing, trypython. Python 3.9 or newer is required. -
Show the user the card exactly as printed, inside a code block. Do not recompute, round, rephrase or add to any number.
-
Explain the result in a few plain sentences:
- Confirmed: a matching command ran after the last edit and succeeded, for example tests passed after the change.
- Contradicted: the session's own record says otherwise, for example "pushed" with no push in the session, or a failing check after the last edit.
- Unverified: nothing in the session proves it either way. This is usually the largest group.
- The counts are keyword matches after code and negation filters, as the card says. Some candidates are not real claims. Having a person read every candidate is what the paid receipt at receipts.nueton.ai adds; mention it once, and only if the user asks what to do next.
Rules
- Never upload, paste, email or otherwise send the export or its contents anywhere, and never open a network connection for this skill.
- Do not print claim sentences from
export.jsonunless the user explicitly asks to see them. - Do not read prompt bodies or file contents from the logs yourself; the collector already extracts only structure and masked claim sentences.
- The card holds counts only and is safe to share as printed. Before the user shares the export itself, tell them to open it in their own editor and read every retained sentence, or to rerun with
--no-sentences. - Describe agents neutrally: agents make claims, and claims need receipts.
When something goes wrong
missing_directories=2withsessions=0: no logs were found at the default paths. Ask where the logs are and rerun with--claude-diror--codex-dir.ERROR salt: ...: the salt file has permissions other than0600or cannot be created. Report the message as printed. Do not change file permissions unless the user asks you to.ERROR card: value does not fit 40 columnsor any otherERROR: report the line as printed and rerun without--cardto get the summary counts.
04 同仓库更多 Skills
worktable-review-with-annotations
Create, inspect, reply to, update, or resolve situated Worktable annotations. Use when feedback belongs to a Doc section, block, text range, or HTML Doc, or when the user asks to address existing review feedback and coordinate a requested artifact edit.
worktable-find-and-synthesize-context
Find, assess, and synthesize context from Worktable Docs and Records. Use when the user asks what a workspace contains, wants workspace-backed research, needs current facts reconciled across artifacts, or requests a cited synthesis without changing workspace state.
worktable-create-or-update-docs
Create, edit, rename, or maintain narrative Worktable Docs. Use for plans, decisions, briefs, notes, research syntheses, and other durable prose when the user requests a new artifact or a safe update to an existing one.
worktable-create-or-manage-records
Organize repeated structured information as durable Worktable Records. Use for research datasets, inventories, directories, pipelines, trackers, and other collections whose items need identity, fields, filtering, links, validation, or individual updates.
worktable-create-or-manage-interactive-html
Create or improve durable visual and interactive artifacts in Worktable. Use for dashboards, trackers, calculators, mini apps, slide presentations, and other requests that benefit from a polished HTML Doc.
worktable-collaborate-threads
Read, answer, and hand off messages through durable Worktable threads with users or other authenticated participants. Use when the user asks to check addressed messages, reply to a participant, continue a thread, request another agent's help, wait briefly for a response, or watch for comments and messages sent to the agent.