agentleFS
Sign inStart free
aliraza2025aliraza2025/receipts/skills/claim-receipts/SKILL.md

claim-receipts

Skill3 starsChanged 5 days agoMITView on GitHub

Audits product writing (READMEs, launch pages, decks, case studies, resumes, PRDs, investor updates) for claims that are not backed by evidence, labels each claim, and rewrites unsupported ones. Use when the user asks to fact-check, review, tighten, or "make honest" any product or marketing copy, or before publishing anything with numbers or outcomes in it.

The file

We split this skill into the rules it gives an agent

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

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    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. Distinctive:

    Every claim in product writing should be able to show its receipt: where it came from and what it covers.

    line 8
  2. Very specific:

    Rewrite: every MEASURED-WITH-SCOPE, ASSUMPTION, and UNSUPPORTED claim.

    line 23

≈ “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

Claim receipts

Every claim in product writing should be able to show its receipt: where it came from and what it covers. This skill finds the claims that cannot, and fixes the wording instead of inventing support.

Workflow

  1. Extract claims. List every sentence that states a fact, number, outcome, comparison, or capability. Skip opinions that are clearly framed as opinions.
  2. Find the receipt. For each claim, look for support in the files, links, data, or notes the user provided. Do not search your own memory for support and do not assume.
  3. Label each claim with exactly one label:
LabelMeaning
SOURCEDA provided source states this directly. Cite it.
MEASURED-WITH-SCOPEA measurement exists, but the claim needs its scope attached (sample, date, conditions).
ASSUMPTIONA plan, estimate, or belief. Fine to keep if worded as one.
UNSUPPORTEDNo receipt found. Rewrite, ask for the source, or cut.
  1. Rewrite every MEASURED-WITH-SCOPE, ASSUMPTION, and UNSUPPORTED claim. Rewrites may only remove or narrow. Never add a number, customer, outcome, or adjective that was not in a source.
  2. Report using the template below.

Red flags to check first

  • Outcomes stated as results when they come from a simulation, demo data, or a plan.
  • Percentages without a denominator, or a denominator without a date.
  • "Users", "customers", or "teams" when the source shows testers, a pilot, or one client.
  • Process metrics presented as quality metrics (jobs completed is not answers correct).
  • Superlatives: first, only, best, leading, revolutionary, 10x.
  • Credit drift: "I built" when the source shows a team, or "we" hiding who did what.
  • Model output presented as verified fact.

Output template

## Claim receipts: [document name]

| # | Claim (quoted) | Label | Receipt or gap | Suggested rewrite |
|---|---|---|---|---|
| 1 | "..." | SOURCED | [source, section] | keep |
| 2 | "..." | UNSUPPORTED | no source provided | "..." |

**Summary:** N claims checked. X sourced, Y need scope, Z assumptions, W unsupported.
**Questions for the author:** sources that would let an UNSUPPORTED claim stay.

Example

Input sentence (a deliberately bad draft): "CrediClip analyzed 780 videos with 97% reliability."

Receipt available: a batch summary showing all 780 jobs completed (two after retry), and 759 of 780 results flagged as low evidence.

Output row:

| 1 | "analyzed 780 videos with 97% reliability" | MEASURED-WITH-SCOPE | Batch summary: 780 of 780 jobs completed; 759 of 780 flagged low evidence. "97%" matches no reported figure. | "In one 780-link YouTube batch, every job completed, but 759 of 780 results were flagged as low evidence." |

The rewrite is less flattering and more useful: it tells the reader what the system did and where it still falls short.

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