v0.8.2

What's changed since v0.8.1

• [issue-174] fix(registry): fill the interface metadata the scanner flags (#175) by @vinicq
• [issue-168] fix(protocol): route every language through the semantic catalog (#168) (#169) by @vinicq
• [issue-166] chore(codex): meet the awesome-codex-plugins listing requirements (#167) by @vinicq
• chore(docs): credit @homesellerq-coder (#160) by @github-actions[bot]
• [issue-158] fix(codex): correct auth enum and document persistent per-project install (0.8.2) (#159) by @homesellerq-coder

🔨 Maintenance

• chore(deps-ci): bump actions/checkout from 7.0.0 to 7.0.1 (#165) by @dependabot[bot]
• chore(deps-ci): bump actions/setup-node from 6.4.0 to 7.0.0 (#164) by @dependabot[bot]
• chore(deps-ci): bump hashgraph-online/ai-plugin-scanner-action from 1.2.464 to 1.2.514 (#163) by @dependabot[bot]
• chore(deps-ci): bump release-drafter/release-drafter from 7.5.1 to 7.6.0 (#162) by @dependabot[bot]
• chore(deps-ci): bump hashgraph-online/ai-plugin-scanner-action from 1.2.392 to 1.2.464 (#161) by @dependabot[bot]

Full changelog: v0.8.1...v0.8.2

v0.8.1

Fixed

• Codex install docs. The .agents/plugins/marketplace.json catalog uses a repo-root source ("path": "./"), which Codex's marketplace resolver does not accept (it expects the plugin nested in a subdirectory of the marketplace root). Stopped advertising codex plugin marketplace add as an install path for this repo and documented the supported path - clone the repo and let AGENTS.md auto-load - across README, docs/user-guide.md, docs/packaging.md, and contexts/codex.md. The plugin command, where shown, is now qualified: codex plugin add falsegreen-skill@falsegreen. Raised by the automated Codex review on #149.

v0.8.0

Added

• Antigravity CLI (agy) support - Google's successor to the discontinued Gemini CLI. Installable plugin (.antigravity-plugin/, agy plugin install) and a workspace Agent Skill (.agents/skills/), both exposed as /falsegreen-skill. Self-contained dist/antigravity-plugin/ build target. agy auto-parses AGENTS.md/GEMINI.md as workspace rule files.

Changed

• Docs present the Antigravity CLI in place of the discontinued Gemini CLI (Gemini provider/API paths - AI Studio, Gemini API, Vertex, --provider gemini - unchanged; legacy artifacts kept as the agy plugin import gemini source).
• Unified the skill name to falsegreen-skill across every host (skills/falsegreen-llm/ → skills/falsegreen-skill/).
• Aligned per-host model recommendations to models.yaml.

Fixed

• Codex marketplace policy enums (AVAILABLE/ON_INSTALL) and dropped the non-schema owner.
• contexts/codex.md plugin-subcommand claim (codex plugin marketplace add).
• Stale CLI default-model docs synced to the code (claude-sonnet-5/gpt-5/gemini-2.5-flash).

Full changelog: CHANGELOG.md.

v0.7.0

Minor release: authoring mode (Mode B) comes to the CLI.

Added

• generate command - render a language-neutral test-spec into one stack (--lang python|typescript|javascript|tsx|jsx|robot), then run Mode A on the result. Fail-closed exits: 0 PASSED, 1 FAILED (surviving false-green), 3 UNVERIFIED. The oracle is mandatory; the self-check is a same-model static review bounded to one revision.
• Docs: README generate section, docs/user-guide.md (three-mode walkthrough, authoring tests for your own feature across the JS/TS family, login vs API-key paths, Claude/Codex/Gemini examples), docs/cli.md provider table, docs/invocation-methods.md, ADR 0004.

Published to npm: npm install falsegreen-skill@0.7.0. Full notes in CHANGELOG.md.

v0.6.3

Documentation: a complete setup and usage reference in the README covering every mode - CLI install, the full analyze + fix flag tables, provider configuration (anthropic/openai/gemini/openai-compatible with base-url + env keys, incl. reasoning-model endpoints like Nvidia NIM, Fireworks, Groq), per-host enable steps (Claude Code, Codex, Gemini, Cursor, plain LLM/API), and the J1-J6 output schema. No detector change.

v0.6.2

Security hygiene: the JWT example-fixture literals that tripped a HARDCODED_SECRET scanner are renamed to obvious placeholders, a package-lock.json is added, and all GitHub Actions are pinned to commit SHAs. This clears the awesome-ai-plugins HOL plugin-scanner gate (72/142 grade C -> 90/142 grade A, 0 high). No detector or protocol change. See CHANGELOG.md.

v0.6.1

Completes reference.md as a true superset of the three static scanners and closes the coverage gap (#105). Adds the missing Python codes C56 (sync assert of a never-awaited coroutine), C57 (compare against an unconfigured Mock attribute), C59 (bare top-level comparison), and PL1 (asserts stripped by -O); documents the Robot C31/C11a/M2 same-id reuse. New guard: schema/scanner-codes.json + scripts/check-scanner-coverage.mjs (wired into validate) fails CI if a scanner ships a code the catalog lacks. See CHANGELOG.md.

v0.6.0

New opt-in AI-fix path (#1): falsegreen-skill fix <test> --case <code> --line <n> --sut <file> proposes a test-only patch for a false-green finding and PROVES it with a mutation gate (parse + behaviour-preserve + a built-in line-scoped mutation the patched test must fail on) before reporting. Propose-only - never auto-applies, never touches the SUT. V1 is Python/pytest, mechanical findings; the honest limit (proves the fix catches the targeted mutant, not every bug) is stated. See CHANGELOG.md.

v0.5.2

CLI reliability fix (#102): --json now works with reasoning models (Qwen3.5, Kimi-K2.6, DeepSeek-style) over the openai-compatible provider. Sends response_format json_object on the OpenAI-style path (verified safe on Nvidia, Fireworks, Groq, Cerebras), hardens JSON extraction (strips think/reasoning, fence + balanced-brace recovery, reasoning_content fallback, key normalization, truncation hint), and fixes a Windows node crash on the error path. See CHANGELOG.md.

v0.5.1

Precision tightenings from real-world field validation: the skill was applied to 200 real repos (50 each Python/TS/JS/Robot, #99) and held up - the gaps were implicit look-alike exemptions, now explicit so an LLM does not over-fire. reference.md adds: non-assert oracles (helper/fluent matchers, pytest runner results, throwing Testing-Library queries, hand-rolled throw/assert, type-level assertions), a sharper mock-vs-SUT boundary (S12/S16/S5), guard exemptions (expect.assertions, pytest.raises message), workshop fixtures, and a contract-scoped Robot file-role rule. SKILL.md generalizes the finding header to {code} ({J}) and adds the C9/C28/S17 tie-break. No new or removed codes (121). See CHANGELOG.md.