Modernize Legends GitHub for portable, evidence-backed repository work

.github/workflows/ci.yml

@@ -21,3 +21,27 @@ jobs:
             SECURITY.md
             SUPPORT.md
             CHANGELOG.md
+            AGENTS.md
+            docs/*.md
+            GEMINI.md
+            github/SKILL.md
+            skills/*/SKILL.md
+            agents/*.md
+            github/references/portable-workflows.md
+            github/references/banner-generation.md
+            github/references/legends-banner-style.md
+            github/references/github-seo-guide.md
+
+  portable-entrypoint:
+    strategy:
+      matrix:
+        os: [ubuntu-latest, windows-latest]
+        python: ['3.10', '3.12']
+    runs-on: ${{ matrix.os }}
+    steps:
+      - uses: actions/checkout@v6
+      - uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python }}
+      - run: python -m pip install -r github/requirements.txt
+      - run: python -m unittest discover -s tests -v

.gitignore

@@ -1,5 +1,7 @@
 # Environment variables (contains API credentials)
 .env
+.env.local
+.env.*.local
 
 # OS files
 .DS_Store

AGENTS.md

@@ -0,0 +1,27 @@
+# legends-github agent entry point
+
+Use this workflow with any agent that can read files and execute commands. No particular model, native skill loader, or subagent framework is required.
+
+## Work on a target repository
+
+1. Resolve the toolkit and target as separate absolute paths. Read the target's instructions and check its Git status. Preserve existing user work.
+2. Establish the user's objective, audience, repository type, and authorized changes. Inspect the actual implementation before recommending templates.
+3. Run `python <toolkit>/legends_github.py verify --mode portable --path <target>`. All readiness modes check the portable runtime without requiring a host installation. Use `--offline --artifacts-dir <outside-target>` for an isolated local review. Optional provider failures do not block unrelated local work.
+4. Run `python <toolkit>/legends_github.py audit --path <target>`, then inspect the evidence and proposed actions. These commands write local caches and reports. An audit score is a checklist, not proof of quality or ranking potential.
+5. Choose the smallest relevant workflow using its `--help`. Distinguish observed facts, unavailable evidence, and hypotheses. Do not promote checklist points directly into business priorities.
+6. Make authorized changes, inspect the diff, and run relevant checks. External mutations, paid services, and publishing must follow the user's actual authorization. Do not repeat approval requests for already authorized work.
+7. Report concrete changes, verification, and remaining limitations. Do not promise higher rankings, stars, revenue, or universal host compatibility.
+
+## Optional instructions and tools
+
+Read specialized `skills/github-*/SKILL.md` instructions when useful. Legacy host-specific paths, mandatory multi-agent language, and optional-provider setup prompts are not requirements of this portable entry point. Translate supported operations into the current host's tools; run sequentially when delegation is unavailable or unnecessary.
+
+Use configured providers only for requested capabilities. Never print credentials or load arbitrary dotenv contents into a shell command. Keep artwork optional and follow the repository owner's design direction.
+
+## Contributing to this toolkit
+
+Keep executable behavior in `github/scripts/`, with `legends_github.py` as the public launcher. Add regression tests for behavior changes. Keep host adapters thin and document their tested status. Do not remove legacy installers or change cache formats without a migration plan. Follow `docs/MODERNIZATION.md` for reconciliation scope.
+
+## DataForSEO research dependency
+
+For measured demand and SERP research, install `requirements-dataforseo.txt` and use `research` to estimate before explicit execution. This uses the public legends-dataforseo-kit; no MCP installation is required. Follow `docs/SEO-RESEARCH.md`, preserve locale and observation dates, and import collected exports through `seo`. Offline audits remain independent of provider credentials.

CHANGELOG.md

@@ -9,11 +9,21 @@ and this project adheres to [Semantic Versioning](https://semver.org/).
 
 ### Added
 
+- Agent-neutral command launcher, capability discovery, offline execution, and per-target artifact isolation.
+- Versioned evidence findings, repository profiles, coverage, and practical priorities alongside the legacy score.
+- Organic discovery plans with sourced local evidence, comparison briefs, experiments, and measurement templates.
+- Regression coverage and Windows/Linux Python CI; read-only GeoGrid validation example.
+
 - Gemini CLI extension manifest and `GEMINI.md` context file so the skill suite can install through `gemini extensions install`.
 - Legends GitHub social preview image.
 
 ### Changed
 
+- Removed the paid image-provider runtime and credential setup. Artwork is optional and prepared from supplied local files.
+- Reworked skills and specialist reviewers around portable execution, evidence, and user outcomes.
+- Simplified the README, removed mascot artwork, and replaced unsupported SEO claims.
+- Fixed audit cache side effects, atomic JSON writes, and metadata topic preservation.
+
 - Renamed public project branding from provider-specific GitHub variants to Legends GitHub.
 - Updated README, citation metadata, support links, installer splash text, and generated artwork for the Legends GitHub name.
 

CITATION.cff

@@ -22,4 +22,4 @@ abstract: >
   Code, Codex, and Gemini CLI. Includes 8 specialized sub-skills covering README
   generation, legal compliance, metadata optimization, SEO keyword research,
   community health, release management, and portfolio strategy. Powered by
-  DataForSEO for live keyword data and KIE.ai GPT Image 2 for generated visuals.
+  optional research integrations for keyword evidence and local artwork supplied by the user.

CONTRIBUTING.md

@@ -25,7 +25,7 @@ Common questions to check first:
 
 - **DataForSEO not working?** Verify your MCP config in `~/.claude/settings.json`. The installer sets this up, but credentials expire or get misconfigured.
 - **Skills not showing up?** Restart Claude Code after installation. Skills register on startup.
-- **Banner generation failing?** Check that `KIE_API_KEY` is set in `~/.claude/skills/github/.env`.
+- **Artwork missing?** Supply an existing local asset or use your host's image tool; a banner is optional.
 
 ## Contributing Code
 

GEMINI.md

@@ -1,18 +1,7 @@
-# Legends GitHub
+# legends-github
 
-Legends GitHub exposes GitHub repository optimization workflows as Gemini CLI agent skills.
+Use `AGENTS.md` as the shared workflow contract. Gemini is one optional host for the portable file and command interface.
 
-Use the bundled skills from `skills/`:
+The bundled specialized instructions live under `skills/`. The source-checkout command entry point is `python legends_github.py`; use `capabilities` to inspect the workflow contract and `--help` for usage.
 
-- `github-audit`
-- `github-legal`
-- `github-community`
-- `github-release`
-- `github-seo`
-- `github-meta`
-- `github-readme`
-- `github-empire`
-
-Run these workflows from the root of the repository you want to optimize. The skills use local source files, git history, GitHub CLI metadata, and optional DataForSEO/KIE.ai credentials to make recommendations.
-
-Follow the GARE pattern in each workflow: gather data, analyze the current state, recommend changes, and execute only after the user approves live GitHub or filesystem mutations.
+Read repository evidence, propose useful changes, and act within the user's authorization. Optional research providers and native image tools are not prerequisites. Run sequentially when delegation is unnecessary or unsupported.