.coderabbit.yaml

# yaml-language-server: $schema=https://coderabbit.ai/integrations/schema.v2.json
# CodeRabbit configuration for fable-mode — ASSERTIVE + MOST STRICT + ADVANCED
# https://docs.coderabbit.ai/reference/configuration
# Profiles available: quiet | chill | assertive  → we lock to assertive (max feedback)

language: en-US

tone_instructions: >-
  You are a principal systems engineer reviewing a deterministic System-2 cognitive
  engine (fable-mode / fable-engine). Be ruthless, precise, and uncompromising.
  Flag every violation of invariants, time-lock semantics, CAS integrity, epistemic
  hygiene, subagent isolation, or concurrency safety. Prefer formal reasoning over
  style nitpicks only when the style itself threatens correctness or maintainability.
  Never soft-pedal; request changes when invariants are at risk.

early_access: false
enable_free_tier: true

reviews:
  # Most comprehensive / nitpicky profile available
  profile: assertive

  # Strictest workflow: CodeRabbit can request changes and block merge until resolved
  request_changes_workflow: true

  high_level_summary: true
  high_level_summary_in_walkthrough: true
  high_level_summary_placeholder: "@coderabbitai summary"
  high_level_summary_instructions: |
    Produce a high-signal summary that explicitly calls out:
    1. Impact on Mechanical Time-Lock / authority deadline semantics
    2. Any change to Main-Agent vs Subagent cognitive separation
    3. CAS / token-compression / FableGrammar333 integrity risks
    4. Epistemic grounding ([PROVEN]/[HYPOTHESIS]/[UNKNOWN]) violations
    5. Concurrency, lock-freedom, or session WAL hazards

  auto_title_placeholder: "@coderabbitai"
  auto_title_instructions: "Use conventional commits: <type>[optional scope]: <description>. Prefer feat/fix/refactor/docs/test/ci for this cognitive engine."

  review_status: true
  review_details: true
  review_progress: true
  commit_status: true
  fail_commit_status: true   # Fail the check on review errors — no silent green

  collapse_walkthrough: true
  changed_files_summary: true
  sequence_diagrams: true
  estimate_code_review_effort: true
  assess_linked_issues: true
  related_issues: true
  related_prs: true

  suggested_labels: true
  auto_apply_labels: false
  suggested_reviewers: true
  auto_assign_reviewers: false

  poem: false
  in_progress_fortune: false
  enable_prompt_for_ai_agents: true

  # Anti-slop for public repo hygiene
  slop_detection:
    enabled: true
    include_all_authors: false
    label: "slop"

  path_filters:
    - "!**/__pycache__/**"
    - "!**/.pytest_cache/**"
    - "!**/*.pyc"
    - "!**/dist/**"
    - "!**/build/**"
    - "!**/.eggs/**"
    - "!**/*.egg-info/**"
    - "!assets/**"
    - "!marketing/**"

  path_instructions:
    - path: "fable_engine/**/*.py"
      instructions: |
        CRITICAL domain: fable-engine is the Mechanical Time-Lock + Session MCP core.
        Enforce:
        - Immutable authority deadline must never be bypassable (hard error on premature unlock_execution).
        - Main Agent MUST NOT write project code; only subagents may mutate workspace files.
        - All epistemic claims must be classified [PROVEN] / [HYPOTHESIS] / [UNKNOWN] with evidence pointers.
        - CAS writes must be atomic (tmp + rename) + SHA-256 verified; never partial or racy.
        - Session WAL / checkpoint_session must remain crash-safe and monotonic.
        - Concurrency: prefer lock-free or carefully ordered locking; document memory model assumptions.
        Flag any weakening of these invariants as blocking.

    - path: "fable_v2/**/*.py"
      instructions: |
        fable_v2 houses the System-3 meta-cognition, proof engine, coder fleet, and design engine.
        Enforce:
        - Subagent contracts (compile_delegation_contract) must remain complete and non-bypassable.
        - Red-team / property-oracle / mutation testing paths must stay adversarial.
        - Design engine (domain: "design") must preserve 7-layer optical depth, Haute archetypes, and anti-slop rules.
        - Vector / plasticity / causal (do-calculus) modules: formal correctness over convenience.
        Request changes on any regression of verification tiers or TRIZ contradiction handling.

    - path: "fable_mode/**/*.py"
      instructions: |
        Installer / launcher / adapters layer. Keep it portable, zero-dependency where possible,
        and never leak secrets or weaken safety.py guards. Prefer pure functions and explicit error paths.

    - path: "tests/**/*.py"
      instructions: |
        Tests must exercise time-lock rejection, CAS integrity, session recovery, and subagent isolation.
        Prefer property-based / adversarial tests. Flag missing coverage of authority-deadline paths
        or epistemic misclassification as high severity.

    - path: "**/*.md"
      instructions: |
        Documentation must stay consistent with the 10 Core Pillars and 6-Phase lifecycle.
        Flag any claim that contradicts the mechanical time-lock or cognitive-separation rules.

    - path: ".github/workflows/**"
      instructions: |
        CI must not skip the mechanical guarantees. Prefer fail-closed behavior.

  auto_review:
    enabled: true
    auto_incremental_review: true
    auto_pause_after_reviewed_commits: 0   # never auto-pause — keep the pressure on
    drafts: false
    ignore_title_keywords:
      - "WIP"
      - "[WIP]"
      - "do-not-merge"
      - "work-in-progress"
      - "skip review"
    labels:
      - "!wip"
      - "!do-not-review"
    base_branches: []   # default branch + any configured

  finishing_touches:
    docstrings:
      enabled: true
    unit_tests:
      enabled: true

  tools:
    ruff:
      enabled: true
    gitleaks:
      enabled: true
    yamllint:
      enabled: true
    markdownlint:
      enabled: true
    shellcheck:
      enabled: true
    github-checks:
      timeout_ms: 600000

chat:
  auto_reply: true
  art: false

knowledge_base:
  opt_out: false