faffter-noon-prd

Name: faffter-noon-prd
Availability: InStock
Author: shftwst

Evaluates PRD stop-condition verifiability and emits a prd-readiness verdict contract block.

3|Updated Apr 16, 2026

Search Tags:#requirements #validation #prd #readiness #gate #verdict #autonomous delivery

Ready to run

npx skills add https://github.com/shftwst/faff --skill faffter-noon-prd-shftwst

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: faffter-noon-prd
Source: https://github.com/shftwst/faff/tree/main/plugin/skills/faffter-noon-prd
Command: npx skills add https://github.com/shftwst/faff --skill faffter-noon-prd-shftwst

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Unattended (L4) delivery runs need a machine-verifiable done-signal before they can start; a vague PRD would let an autonomous run loop forever against prose goals. This Skill judges whether a PRD's stop-conditions are concrete enough to admit the run. ## Core Features & Use Cases - PRD admissibility judgement: Reads only the PRD document and decides whether its done-criteria are machine-checkable (Given/When/Then scenarios or single MUST/comparator assertions). - Creative-licence classification: Labels the PRD's implementation latitude as `broad` or `tight` to calibrate downstream scope strictness. - Contract-based verdict output: Emits exactly one `faff-contract:prd-readiness` JSON block that the deterministic `faff contract prd-readiness` validator checks to admit or refuse the run. - Use Case: Before an L4 lights-out run starts, the run-start gate pipes this Skill's verdict block to the validator; a PRD with only vague goals like "the system should be stable" is refused with a `not-ready` verdict. ## Quick Start Ask the agent to evaluate the PRD at the given absolute file path and produce the prd-readiness verdict block for the run-start gate.

Frequently Asked Questions about faffter-noon-prd

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I check if a PRD is ready for an unattended autonomous run?▼

PRD readiness is checked by judging whether its done-criteria are machine-verifiable, such as Given/When/Then scenarios or single MUST/comparator assertions. The skill emits a prd-readiness verdict block that the deterministic faff contract prd-readiness validator uses to admit or refuse the run.

What makes PRD stop-conditions machine-verifiable?▼

A stop-condition is verifiable when written as a Given/When/Then scenario or a single MUST/comparator assertion. Loose prose like "the system should be stable" or "users will be satisfied" is not verifiable, and at least one verifiable condition is required for an admissible verdict.

Does the PRD readiness check look at the codebase or tracker?▼

No, the check is code-blind by construction and reads only the PRD document at the given absolute path. It never inspects the codebase, tracker, run context, or specs, so the verdict describes the PRD alone and cannot be influenced by plausible code.

What happens when a PRD has vague or missing done-criteria?▼

The verdict is not-ready with a non-empty reason from the closed reason set, and the gate refuses the run. The gate is fail-safe toward refusal, so a malformed or missing verdict block never coerces toward admissible.

What is the creative_licence field in the prd-readiness verdict?▼

creative_licence classifies the PRD as broad or tight, describing how much implementation latitude the team has. It is forward-carried to calibrate downstream scope strictness but is never gate-decisive for the admit/refuse decision.