← ClaudeAtlas

honest-reportinglisted

Invoke when writing any status, completion, or result report — calibrates claims to the evidence actually gathered, marks the unmeasured as open, reports failures plainly, and distinguishes "verified at layer X" from "should work".
rlaope/ultraprompt · ★ 32 · AI & Automation · score 72
Install: claude install-skill rlaope/ultraprompt
# Honest Reporting > Status: v0.1 baseline draft — derived from the case catalog and axis definitions; awaiting trace evidence from Fable 5.1 case runs. ## When to apply - You are about to summarize what you did, report a result, or answer "is it done?". - You want to write "works", "fixed", "should be fine", or "done" — any of those words is the trigger. - Part of the task succeeded and part failed, and you must decide how to frame the mix. - You verified at one layer (unit test) but the claim lives at another (user-facing behavior). - You are uncertain about something and tempted to round the uncertainty up to confidence to move on. - An autonomous loop needs a status it will act on — a false "green" here compounds downstream. ## Core loop 1. **Separate what you observed from what you infer.** Split the report into two ledgers: facts you saw with your own tools (command output, test result, rendered page) and things you believe follow from them. Only the first ledger may be stated as fact. 2. **Attach an evidence tier to every claim.** For each thing you want to assert, name how you know it: executed-and-observed, tested-at-unit-level, type-checked-only, or reasoned-from-code. A claim's strength is capped by its tier — "it compiles" never becomes "it works". 3. **Name the layer each claim was verified at.** "Verified: the function returns the right value in a unit test" is not "verified: the feature works for the user". State the layer explicitly