Skip to content

Investigation: Disposition of Upstream Gherkin v2 Experiment Findings (#434)

Date: 2026-04-28 Scope: Investigation only — no code changes Disposition: Closed as not-planned


Summary

Issue #434 asked whether to port the upstream gherkin_v2_experiment_findings.md document into amplihack-rs, or remove the "gherkin-expert" string from known_skills.rs if the feature is unused.

Conclusion: CLOSE AS NOT-PLANNED — no parity gap exists.

The gherkin-expert capability already ships in amplihack-rs through the amplifier-bundle. The upstream document is an informational experiment log, not a feature specification. There is nothing to port and nothing to remove.


1. What the Issue Claimed

Issue #434 (tracked from the #420 documentation parity audit) reported:

  • Upstream carries docs/gherkin_v2_experiment_findings.md
  • The only amplihack-rs hit for "gherkin" is a string in known_skills.rs
  • No agent, recipe, or workflow uses gherkin v2 today

The issue suggested either porting the experiment as an eval scenario or removing the string from known_skills.rs.

2. What the Codebase Actually Contains

A broader search reveals the gherkin-expert is a live, functional skill:

File Purpose
amplifier-bundle/skills/gherkin-expert/SKILL.md Full skill definition with triggers, usage, and examples
amplifier-bundle/agents/specialized/gherkin-expert.md Agent definition for BDD/Gherkin test generation
crates/amplihack-hooks/src/known_skills.rs Valid allowlist entry referencing the bundle skill
docs/howto/use_gherkin_expert.md User-facing how-to guide
docs/guides/formal-specifications-as-prompt-language.md Related guide referencing gherkin patterns

The initial grep (crates/*/src/) only searched Rust source files, missing the bundle and documentation directories where the skill lives.

3. Why No Action Is Needed

The upstream document is informational, not a feature spec

gherkin_v2_experiment_findings.md records the results of a past experiment. It describes what was tried and what was learned. It is not a specification for a feature that needs implementing.

The gherkin-expert capability already exists

The amplifier-bundle ships a complete gherkin-expert skill (SKILL.md) and agent (gherkin-expert.md). The known_skills.rs entry correctly references this bundle skill. Removing it would break skill discovery.

No parity gap

The parity audit flagged this because the grep was too narrow. The capability exists — it is bundled, documented, and registered.

4. Decision

Aspect Decision
Port upstream document? No — it is an experiment log, not a feature spec
Remove known_skills.rs entry? No — it references a valid, live skill
Create eval scenario? No — skill already functions; eval is orthogonal
Issue disposition Closed as not-planned

5. Risks Considered

  • False positive from narrow grep: The original issue was filed based on a grep limited to crates/*/src/. A repo-wide search immediately shows the skill exists. Future parity audits should search the full repository including amplifier-bundle/.

  • Removing a valid entry: Had the known_skills.rs entry been removed, the gherkin-expert skill would no longer be discoverable by the hooks system. This would be a regression, not cleanup.