lint: an asset substituted into a call's argument is used, and is not in a prompt - #18
Merged
Merged
Conversation
… in a prompt
Two rules were wrong about the same reference form, in opposite directions.
A call passes an asset either by reference, `{from: "asset:x"}`, or as text
substituted into an argument, `code: "{{asset:x}}"`. Only the first was
collected. So W20 called live payloads dead — nineteen of them in one live
catalogue, not one real — and a rule wrong that often is worse than a missing
one: it teaches people to skim past the whole report.
Counting the second form as ordinary in-text use would have traded that for the
opposite error: W4 warns that code substituted into text goes through the
model's context, which is true of an instruction and false of a call. A call is
executed by the host and its arguments never enter a prompt. So the two are now
tracked separately: everything counts as "used", only instruction text counts
as "seen by the model".
The fixture for this is CLEAN, which is its job — it stands guard over silence.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Two rules were wrong about the same reference form, in opposite directions.
A call passes an asset either by reference,
{from: "asset:x"}, or as text substituted into an argument,code: "{{asset:x}}". Only the first was collected.So the two are tracked separately now: everything counts as used, only instruction text counts as seen by the model.
Measured on a live catalogue of 35 skills: findings 25 → 9, with no new W4 appearing. The remainder are real (size warnings).
The fixture added here is clean, which is its job — it stands guard over the silence.