Add falsegreen-skill to Community Plugins - #50
Conversation
Test-quality skill that detects false-positive tests: https://github.com/vinicq/falsegreen-skill
|
Warning You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again! |
|
Updated the plugin: the HOL plugin-scanner now reports 90/142 (grade A), 0 high/critical on |
|
Done. Added
The scanner actions are SHA-pinned so the workflow doesn't cost points on its own Operational check. I also linked the run and score in the PR description. Ready for re-review whenever convenient. |
|
🎉 Hey @vinicq, your plugin has been merged and is now listed in the HOL Registry! Claim your pluginAs the author, you can verify ownership of your plugin to unlock:
How to claim
The whole process takes under 30 seconds. No need to add any secrets or tokens to your repo — verification is done entirely through GitHub OAuth. If you have any questions, feel free to ask here or reach out at support@hol.org. |
The entry merged in hashgraph-online/awesome-ai-plugins#50 was removed 11 days later by that repo's sync bot, which overwrites its own README with the one from hashgraph-online/awesome-codex-plugins. The mirror is not a writable surface, so the submission goes upstream, where the validator is stricter. Two hard failures in scripts/validate-plugin-pr.py, both fixed here: interface.composerIcon was absent from the Codex manifest, and the repo had no icon asset at all. The icon is a broken green check, 413 bytes, pure vector, no raster payload, legible at 32px. Also fixes a real host defect: skills/falsegreen-skill/SKILL.md mandated loading reference.md in full for any non-Python file, while contexts/codex.md documents in this same repo that its ~90 KB never fits the ~32 KiB Codex budget and truncates mid-file without warning. That silently under-detects TypeScript, JavaScript and Robot Framework. The load is now section-scoped, with the compact fragments/* tables named as the entry point for budget-constrained hosts. The scanner workflow now matches the filename and the settings that the upstream PR checklist names. min_score was unset, so the action ran with MIN_SCORE: 0 and the score gate was never enforced. The header comment cited the wrong repo and a score scale that no longer exists. Closes #166
|
@kantorcodes heads up, since you merged this one: the entry is no longer in the README here. It was removed on 2026-07-14 by - name: Sync README from codex
run: cp codex-source/README.md README.mdUpstream never had the line, so I resubmitted at the source rather than reopening anything here:
That PR also covers what this repo's guide did not require at the time: Nothing needed on this PR. Leaving the note here so the trail is not confusing for anyone who finds the merge and then cannot find the entry. |
2 participants
Adds one entry to Community Plugins -> Development & Workflow for falsegreen-skill.
It is a test-quality skill for coding agents (Claude Code, Codex, Gemini) that detects false-positive tests: tests that keep passing green even when the code under test is broken. It applies a J1-J6 judgment framework (does the assertion run, does the value come from an independent oracle, is the real unit tested, etc.) across Python, TypeScript, JavaScript, and Robot Framework.
.claude-plugin/,.codex-plugin/, andgemini-extension.json; listed once here as a skill to avoid duplicate entries.Single additive line, inserted alphabetically. One-sentence description per CONTRIBUTING.md.
Scanner CI
The HOL Plugin Scanner now runs in CI on the source repo via a dedicated workflow (
.github/workflows/plugin-scanner.yml), on every push and pull request, failing on any high/critical finding.Scanner actions are SHA-pinned to keep the Operational score clean.