Skip to content

chore(openspec): name-only rf-core dependencies in plugin-split; rf-bdd eval notes - #47

Merged
manykarim merged 2 commits into
mainfrom
chore/plugin-split-spec-and-bdd-evals
Nov 21, 2026
Merged

manykarim merged 2 commits into
mainfrom
chore/plugin-split-spec-and-bdd-evals

Conversation

@manykarim

@manykarim manykarim commented Nov 21, 2026 •

Copy link
Copy Markdown
Collaborator

Summary

  • plugin-split spec: "Technology plugins depend on core" still required a version range of the core major version. Since 5.0.0 every plugin is released at one version from one commit, plugin.json declares "dependencies": ["rf-core"], and build-catalog.py refuses a range. The requirement now says name-only, notes that Copilot CLI, VS Code and Codex install no dependencies (the README tells their users to install rf-core), and adds a scenario for the refused range.
  • add-rf-bdd-skill tasks:
    • 1.1: replaced by the stored GLM trigger baseline, because rf-bdd evals run only on CI with OpenRouter.
    • 5.5: CI results recorded. Validation recall is 4/4, but should-not accuracy is 3/4 ("SpecFlow in C#" loads rf-bdd), below the 0.90 target, accepted by the maintainer: the skills cover Robot Framework only, and a SpecFlow project has no rf-bdd installed. All tasks are done.
    • 7.3: records the merge of feat(rf-core): add the rf-bdd skill #46.

Test plan

  • openspec validate --all --strict (38 passed)
  • check-docs-voice.py
  • CI

AI assistance

Written with Claude Code (Claude Opus 5.5) at the maintainer's request; the maintainer chose the CI-only eval approach.

🤖 Generated with Claude Code

https://claude.ai/code/session_01FHEmzL3ErAhGQykKWDx2wP

manykarim and others added 2 commits November 21, 2026 09:20
…dd eval notes

plugin-split still required a version range on the rf-core dependency.
Since 5.0.0 the dependency is by name only, and build-catalog.py refuses
a range; the requirement now says so, names the agents that install no
dependencies, and gains a scenario for the refused range.

add-rf-bdd-skill: tasks 1.1 and 5.5 are measured on CI with OpenRouter
GLM 5.3 Flash only (maintainer decision). 5.5 records that validation
should-not accuracy is 3/4, below the 0.90 target, and stays open;
7.3 records the merge of PR #46.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FHEmzL3ErAhGQykKWDx2wP
@manykarim
manykarim merged commit ec5ea8b into main Nov 21, 2026
24 checks passed
@manykarim manykarim mentioned this pull request Nov 21, 2026
1 of 2 tasks
@manykarim
manykarim deleted the chore/plugin-split-spec-and-bdd-evals branch November 21, 2026 13:16
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant