deterministic-judgment
/deterministic-judgment
Use when adding, reviewing, or choosing deterministic logic versus model-led reasoning, especially classifiers, validators, rule engines, source selection, legal/tax/compliance workflows, LLM fallbacks, or any change justified by determinism, repeatability, or token savings.
- Author
- fabioannovazzi ↗
- Repository
- app_files ↗
- Licence
- AGPL-3.0
- Last repository change
- 27.10.2026
Assessment and security
- Files in this package
- 1
- Scripts
- 0
- HTML comments
- 0
- URLs in SKILL.md
- 0
No signals detected. Repository changed on 27.10.2026, without references to retired models or APIs.
AGPL-3.0, copyleft: przeróbki udostępniane innym muszą mieć tę samą licencję.
What it does and when to use it
Use when adding, reviewing, or choosing deterministic logic versus model-led reasoning, especially classifiers, validators, rule engines, source selection, legal/tax/compliance workflows, LLM fallbacks, or any change justified by determinism, repeatability, or token savings.
Original author description
Use when adding, reviewing, or choosing deterministic logic versus model-led reasoning, especially classifiers, validators, rule engines, source selection, legal/tax/compliance workflows, LLM fallbacks, or any change justified by determinism, repeatability, or token savings.
How it works
- Required Standard
- Decision Rule
- Implementation Requirements
SKILL.md preview
--- name: deterministic-judgment description: Use when adding, reviewing, or choosing deterministic logic versus model-led reasoning, especially classifiers, validators, rule engines, source selection, legal/tax/compliance workflows, LLM fallbacks, or any change justified by determinism, repeatability, or token savings. --- # Deterministic Judgment Deterministic is not preferred by default. Use deterministic logic only when it is demonstrably better for the specific job than model-led reasoning. ## Required Standard Before adding or preserving deterministic logic, state the reason it should be deterministic. A valid reason must identify at least one concrete advantage: - mechanically verifiable correctness; - lower observed error rate than model-led reasoning on representative cases; - strict reproducibility required by an external contract; - security, safety, or auditability that depends on fixed rules; - performance or cost savings after quality is shown not to regress. Invalid reasons: - deterministic is better in theory; - deterministic saves tokens; - deterministic feels safer; - deterministic makes the system look more engineered; - the task has keywords that can be classified with simple string matching. ## Decision Rule Use deterministic code for mechanical work: parsing stable formats, schema validation, exact arithmetic, file packaging, cache keys, ID matching, simple presence checks, and transformations with explicit contracts. Use model-led reasoning for semantic judgment: legal or tax relevance, source selection, topic taxonomy, research scope, document meaning, claim interpretation, ambiguous intent, and anything where context changes the answer. Hybrid designs are acceptable only when the boundary is clear: deterministic code may collect evidence or enforce output shape, but it must not overrule model judgment on semantic relevance unless tests show it is better. ## Implementation Requirements - Include a short comment, docstring, or PR note explaining why any new deterministic rule is justified. - Add representative tests for deterministic rules, including false-positive and false-negative cases when the rule gates behavior. - For legal, tax, or compliance modules, do not let deterministic classifiers choose legal topics, governing framework, source domains, or research phasing. - If deterministic output disagrees with model judgment on a semantic issue, prefer model judgment unless a documented benchmark or external requirement says otherwise. - Remove deterministic logic that exists only because it is deterministic and has no demonstrated quality advantage. - When reviewing an execution block, trace the rejected condition to the concrete failure it prevents. An existing rejection test establishes current behavior, not that the rejection is necessary. Check a valid case that should proceed as well as the failure case. External version labels and installation history alone do not establish missing runtime capabilities.
Versions
The database has no other skills named deterministic-judgment.
Database snapshot: 13.11.2026 · package version 34f5a05adb8f (SHA-256 digest of the entire package)
Downloads in the last 30 days: 0 · ranking
Report a problem with this skill: harmful content, incorrect descriptions or an author’s removal request.
Similar skills
No similar skills in the database.