CC fixer
The eval-to-repair loop — the consumer of the evaluator's report (#409). Reads the eval output (eval-deep-vision.json's `what_to_redo` first, else eval.json `findings[]`), runs `ralphy project repair-plan <id>` to classify every finding by owner (art-director / scenarist / editor) and order it by severity, presents the ordered plan grouped by owner, gets the user's approval, then applies targeted fixes through the EXISTING `ralphy` verbs (art-director regen / scenarist rewrite / editor recut), re-renders, re-evals, and compares the old vs new verdict. HARD GATE: no paid model regeneration runs until the user approves the plan (or previously opted into batch repair). The deterministic `repair-plan` step itself makes ZERO model calls and is safe to run unprompted. USE WHEN the user says "fix the eval", "repair this render", "make it shippable", "fix the issues / problems / findings", "address the eval", "now fix it", "make the fixes", or hands off a failed / warn eval and asks what's next. ALSO FIRE right after the evaluator returns a `warn` / `fail` verdict and the user wants action. TRIGGER (EN): "fix the eval", "repair the video", "make it shippable", "fix the findings", "apply the fixes", "address the eval issues", "fix what's wrong", "repair plan", "now fix it". See body for ALSO FIRE / DO NOT FIRE / HARD INVARIANTS.
The eval-to-repair loop — the consumer of the evaluator's report (409).
As a process C 59/100 · Has gaps — weak spots: result and completion, inputs and preconditions, running it twice
How to improve
- Shorten the description to 1024 characters.
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 0
✓ No critical or high findings
Files scanned: 1. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- error
description-longdescription is 1341 chars, limit 1024 - note
frontmatter-keyunknown frontmatter key "namespace"
Process rating: all ten parameters 59/100
- 0Result and completion. Does not say what the result is
- 0Inputs and preconditions. Does not say what the process needs to start
- 0Progress reporting. Says nothing while it works
- 30Running it twice. 5 mutating operations with no state check
- 60Tools and files. Uses tools (web) that frontmatter does not declare
- 70When it triggers. States when to use, but not when not to
- 100Steps. 25 steps
- 100Failures and branches. 1 branches, has a failure section
- 100Consistency. Name and required fields are in place
- 100Execution cost. Instruction body is 1870 tokens
- low The skill ranks results itself: that belongs to the system behind the tool, not the model
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +4Description does not say when NOT to use the skill (false activations)
- +3Description length 1341: 120–800 characters recommended
- +3Output format is not stated: the model decides each time
- +1No license
- +2Single-language instructions
- +5Description quotes 16 example trigger phrases
- +4Structure: 13 headings
- +3Step-by-step instructions: 25 items
- +4Has examples (2 code blocks)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 60.