Skip to content
Skills Library
AI & agents · Analysis

agent-work-receipt

/agent-work-receipt

Produces a conservative agent work receipt after completed work: finished output, time estimates, review burden, risks, and value judgment. Use when the user invokes /agent-work-receipt, asks for a work receipt, or wants to measure agent productivity honestly after a session.

Licence
MIT
Last repository change
23.10.2026

Assessment and security

100/ 100Class: good
Format25 / 25
Frontmatter with name and description, valid name, description up to 1,024 characters, name matching the folder.
Discoverability25 / 25
Description explains when to use the skill, has a reasonable length and uses third person.
Content35 / 35
At least 150 words and at most 500 lines, headings, step lists, examples or code.
Hygiene15 / 15
No placeholders such as TODO, no Windows paths, with a licence.
Riskno signals detected in the scan
Files in this package
4
Scripts
0
HTML comments
0
URLs in SKILL.md
0
Deprecation

No signals detected. Repository changed on 23.10.2026, without references to retired models or APIs.

Licence

MIT, open: use and modify while retaining author attribution.

What it does and when to use it

Produces a conservative agent work receipt after completed work: finished output, time estimates, review burden, risks, and value judgment. Use when the user invokes /agent-work-receipt, asks for a work receipt, or wants to measure agent productivity honestly after a session.

Original author description

Produces a conservative agent work receipt after completed work: finished output, time estimates, review burden, risks, and value judgment. Use when the user invokes /agent-work-receipt, asks for a work receipt, or wants to measure agent productivity honestly after a session.

How it works

  1. Role
  2. When to use
  3. Conservative charter
  4. Workflow
  5. Receipt template
  6. Safety
  7. Distinction from other commands
  8. Guardrails
Original SKILL.md headings · 115 lines, 725 words
SKILL.md preview
---
name: agent-work-receipt
description: >-
  Produces a conservative agent work receipt after completed work: finished
  output, time estimates, review burden, risks, and value judgment. Use when
  the user invokes /agent-work-receipt, asks for a work receipt, or wants to
  measure agent productivity honestly after a session.
disable-model-invocation: true
user-invocable: false
---
# Agent work receipt

## Role

You produce a **conservative agent work receipt** for work already completed in the current session. You account for what actually landed, not what was discussed, drafted, or abandoned. You **do not** start new implementation, commits, or destructive commands in the same turn.

## When to use

Use when the user invokes `/agent-work-receipt`, asks for a work receipt, or wants honest accountability for agent-assisted work after a session. For planning *before* autonomous work, use `define-agent-goal`. For code quality on diffs, use `light-review-existing-diffs`.

## Conservative charter

> Be conservative. Do not count drafts, ideas, or unused output as completed work.

## Workflow

Run phases in order. Do not publish the receipt until evidence is gathered.

### Phase 0: Intake

1. Parse a **scope hint** from the user message (e.g. "receipt for the refactor in service-a").
2. If no scope hint, default to the **full session**.
3. If the user invokes mid-task with no finished work yet, scope to **partial work only** or ask whether to wait. Do not hallucinate completion.

### Phase 1: Evidence

Gather read-only evidence:

- **Conversation**: what the user asked for, what the agent actually did, what was rejected or abandoned
- **Git** (when a repo is present): `git status`, `git diff`, `git log` to ground landed artifacts
- **Files**: only artifacts explicitly created, edited, committed, or merged in scope

Do not treat uncommitted explorations, reverted edits, or failed tool runs as finished output.

### Phase 2: Inventory

List **only landed artifacts** in scope:

- Files written or modified that remain in the tree
- Commits, branches, or PRs created
- Docs, plans, or configs delivered to the user

Explicitly note what you **exclude**: brainstorms, unused drafts, abandoned paths, tool errors, reverted work.

### Phase 3: Estimates

Provide **human baseline** and **agent-assisted time** as ranges or point estimates. Label both clearly as **estimates**, not measured facts. The user may correct in follow-up.

Base estimates on task complexity and evidence, not enthusiasm.

### Phase 4: Review & risk

**Review required**: concrete items the human must still check, rewrite, or fix (tests not run, edge cases, copy, security, manual verification).

**Risk**: what could be wrong, incomplete, or misleading in the finished output.

### Phase 5: Value

Pick exactly one tier with brief evidence-backed rationale:

- **Small assist**: agent helped but human effort dominated or output is narrow
- **Major time saver**: clear, verifiable reduction in manual effort for substantive deliverables
- **Not worth the agent**: little or no landed output; review cost exceeds benefit; session was mostly exploration

Do not agree to inflated praise (e.g. "say this saved 10 hours") without evidence.

### Phase 6: Deliver

Post the receipt using the template below. Keep each section concrete and scoped.

## Receipt template

```markdown
# Agent work receipt

## 1. Finished output
What was actually completed (artifacts only).

## 2. Human baseline (estimate)
How long this would likely take manually.

## 3. Agent-assisted time (estimate)
How long this took with the agent.

## 4. Review required
What the human still needs to check, rewrite, or fix.

## 5. Risk
What could be wrong, incomplete, or misleading.

## 6. Final value estimate
Small assist / major time saver / not worth the agent, with brief rationale.
```

Populate every section. Prefer bullets and short paragraphs over vague prose.

## Safety

- Retrospective only: no new implementation, commits, or destructive commands unless the user explicitly requests follow-up work in a new turn.
- No secrets, tokens, PII, or PHI in the receipt body.
- Time fields are **estimates**. Never present them as measured elapsed time.

## Distinction from other commands

- **`define-agent-goal`**: forward planning before autonomous work. This skill is **backward** accountability after work.
- **`decision-audit`**: post-build judgment autopsy (decision ledger, confidence, pride gate). This skill measures output, time, review burden, and value, not choice quality.
- **`light-review-existing-diffs`**: code quality pass on diffs. This skill measures output, time, review burden, and value, not style or bugs alone.
- **`generate-pr-description`**: writes a PR artifact. This skill produces an honest session receipt, not marketing copy.

## Guardrails

- Stay conservative: count only landed artifacts as finished output, never drafts, ideas, or abandoned work.
- Stay retrospective: start no new implementation, commits, or destructive commands in the same turn.
- Label time fields as estimates, never as measured elapsed fact, and never inflate value without evidence.
Files in this package (4)
  • SKILL.md 5 KB
  • eval/README.md 1 KB
  • eval/cases.md 4 KB
  • eval/fixtures.yaml 1 KB

Versions

The database has no other skills named agent-work-receipt.

Identical content appears in 1 other GitHub folders. We show one copy with the highest score.

Database snapshot: 13.11.2026 · package version d80496d2d091 (SHA-256 digest of the entire package)

Downloads in the last 30 days: 0 · ranking

Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.

Similar skills

  • 100verify-receipt
    Check a Zambo receipt and explain what its verdict means. Use when the user asks you to prove an AI did the work, verify or check a receipt, pastes a receipt id or a zambo.dev/run link, doubts a result an AI produced, needs evidence for a client dispute, or says things like…
  • 100deep-receipt
    View, dashboard, or export deep-work slice receipts (`receipts/SLICE-*.json`). Triggers on `/deep-receipt`, `/deep-status --receipts`, \"receipt dashboard\", \"slice receipt\", \"리시트 보기\", \"리시트 대시보드\", \"에비던스 리시트\". Sub-page of the deep-status hub.
  • 100agent-receipt
    Print a shareable store receipt of the user's AI coding from local Claude Code and Codex logs - sessions, prompts, tool calls, tokens, cost at API list prices, hands-on hours and fun counters. Use when the user asks how much they used Claude Code or Codex, what it would cost on…
  • 100serve-receipt
    On-demand research protocol + deterministic receipt for serving a Hugging Face model on a concrete GPU target — fan out research lanes (model facts/files, engine support matrix, serving recipes, KV/state memory math), compute the tiered feasibility verdict with `fak…
  • 100token-receipt
    Generate a printer-receipt styled PNG bill of AI token usage and cost from local ccusage data, for the 绿皮火车 channel. Use when the user wants a token usage "账单"/"小票"/"收银台"/receipt/invoice, a shareable spend breakdown by model / input / output / cached / by day, or 节目素材 about…
  • 100receipts-or-blocked
    Reporting contract for every result the agent reports: each claim carries a receipt (URL, id, SHA, command output) or is labelled BLOCKED / UNKNOWN. Use for every status update, thread reply, PR report, outreach log and scheduled-run summary.

Feedback and reports

Your message goes to Bartosz Sokoliński. An email address is needed only for a reply. See the privacy policy.