classify-spec-strengths

Classify Spec Strengths

Generate

When the user says "review specs for", "generate spec review for", or provides directories/files to analyze, first generate the markdown and JSON:

~/projects/veracity/target/release/veracity-review-module-fn-impls -d src/ChapNN

Or for specific files:

~/projects/veracity/target/release/veracity-review-module-fn-impls -f src/ChapNN/File.rs

Multiple directories can be combined in one run (-d src/Chap05 -d src/Chap18).

Output goes to analyses/veracity-review-module-fn-impls.md and .json.

Classify

When the user says "classify spec strengths", "review spec strengths", "classify specs", or similar:

Inputs

• JSON file: analyses/veracity-review-module-fn-impls.json (or a path the user provides)
• Prompt: ~/projects/veracity/docs/veracity-classify-spec-strengths-prompt.md

Procedure

1. Read ~/projects/veracity/docs/veracity-classify-spec-strengths-prompt.md to understand the classification criteria.
2. Read the JSON file the user specifies (default: analyses/veracity-review-module-fn-impls.json in the current project).
3. For each entry, examine the snippet field and classify spec_strength as one of:
   • strong — requires and ensures fully capture the function's contract.
   • partial — some spec exists but is incomplete or underspecified.
   • weak — spec exists but is trivially true or nearly useless.
   • none — no requires/ensures at all (empty snippet or signature only).
4. Write the classifications to analyses/review-module-fn-impl-spec-strengths.json in the same directory as the input JSON. Format:

   [
     { "id": 1, "spec_strength": "strong" },
     { "id": 2, "spec_strength": "none" }
   ]
   

5. Run the patch command:

   veracity-review-module-fn-impls --patch \
     analyses/veracity-review-module-fn-impls.md \
     analyses/review-module-fn-impl-spec-strengths.json
   

6. Print a summary table:

   Classification	Count
   strong	N
   partial	N
   weak	N
   none	N