[pull] master from flypythoncom:master - #10

Generated with [Devin](https://devin.ai)

Co-Authored-By: Devin <158243242+devin-ai-integration[bot]@users.noreply.github.com>

…ontracts

- courses/ content type with the flagship Claude Code hands-on course
  (EN+ZH lessons, three scenario skins, task contract, starter/solution,
  verify.py objective evidence, REVIEW.md record)
- tools/verify_courses.py course-contract verifier wired into Makefile and CI
- catalog/projects/ per-project YAML (7 seeds) with ai_familiarity,
  bilingual generated Radar tables, radar.json export + radar-v1 schema,
  and --target both support in export_catalog.py
- tools/radar_scan.py read-only candidate discovery (GitHub/PyPI/HN) with
  the no-verdict guarantee enforced by tests
- content-manifest schema + builder gain the course type
- course-feedback issue template; llms.txt courses and radar sections
- README course banner plus contextual flypython.com footers on every
  guide, playbook, and example per REPO_TO_WEBSITE.md

C2 'Hands-on with OpenAI Codex CLI' ships EN+ZH lessons, the same three
scenario skins and the same verified starter/solution/tests contract as C1
per the 0.0.3 plan reuse rule; REVIEW.md records the mechanical verification
with the live Codex run-through logged as pending launch evidence.

GitHub moved jxnl/instructor to 567-labs/instructor; the radar entry and the
generated radar.json/README tables now use the canonical URL so the site's
external-link gate stays green.

- C3 'One source of truth for agent rules': new stdlib rules_check.py core
  (source/pointer/copy/drift classification + CLI exit codes, 11 tests)
- C4 'From it runs to it ships': new stdlib ship_check.py core (parsed test
  counts, no-tests verdict, unverified list, atomic SHIP-RECORD.json,
  10 tests) with green/red/no-tests scenario projects
- C5 'Give your agent tools with MCP': assembled from the reviewed
  examples/mcp-server contract plus wire-format request samples
- All five courses EN+ZH with COURSE teaching contracts, scenario skins,
  TASK contracts, REVIEW records, and verify.py both-directions evidence

Planning draft for the local-first challenge/badge layer in course folders:
verify.py progress writing PROGRESS.json, badge contract and challenge
narrative (EN+ZH), gated BADGE.md artifact, SKILL.md packaging evaluation,
and verify_courses.py coverage. Carries the open 0.0.3 release mechanics
forward.

Rev 2 superseding the local-first draft: the website records progress from
deterministic checkpoint claim codes printed by verify.py (FP-411), the
challenge/badge narrative points at the site recording flow (FP-412), badge
metadata becomes structured course data (FP-413), and verify_courses.py
gains claim-code determinism coverage (FP-415). Local-only artifacts are
dropped.

Owner decision 2026-09-12: drop the English+Chinese doc pairs —
repo-plan-0.0.3 and repo-plan-0.0.4 each become one Chinese-only
document, and AGENTS.md records that plans are maintained in Chinese
while other governance docs stay English-canonical.

M0 intro module (agent tool ecosystem overview), the dual-agent route
challenge, and paths/foundation/ contract reusing the live C1-C4 courses
as modules 1-4; full path system defined in the site's LEARNING-PATHS.md.

- Claim code = flag contract; serves local challenge loop first,
  website server recording later
- FP-412 reframed: learner solves challenges; COURSE.md becomes
  optional guided-mode entry
- Agent-solvability is a launch gate for every challenge
- No repo work pairs with website phase B until catalog depth met

- Course folder retrofit: TASK.md as challenge entry, COURSE.md as
  optional guided mode, badge frontmatter contract, claim-code
  derivation (salt + course_id + checkpoint_id + evidence_hash)
- paths/ layer: PATH.md structured contract, capstone challenge,
  points table (10/50/200)
- Contract follow-ups: manifest type 'path', curation scope, README
  narrative, solvability records in REVIEW.md
- Landing order: C1 pilot -> C2-C5 rollout -> foundation path ->
  DA courses

Merges feat/0.0.3-courses-and-radar: the five-course batch (C1-C5,
EN+ZH), per-project Radar YAML with radar.json + schema, claim-code
 groundwork plan (0.0.4 docs), README referral, and the instructor
canonicalization.

ubuntu-latest images now ship an externally-managed system Python
(PEP 668), so 'uv pip install --system' refuses to install and the
Validate workflow dies at the dependency step. Create the project venv
from the matrix interpreter and run every gate through 'uv run'.

'uv run' re-syncs the project environment and rebuilds it on the system
interpreter, dropping uv-pip-installed tools like ruff. Invoke the venv
binaries directly so the matrix interpreter and installed tools are
exactly what runs.