apple-a-day 0.2.0__tar.gz

apple_a_day-0.2.0/.github/ISSUE_TEMPLATE/bug_report.md

@@ -0,0 +1,20 @@
+---
+name: Bug report
+about: Something apple-a-day got wrong or missed
+labels: bug
+---
+
+**What happened**
+A clear description of the bug.
+
+**Expected behavior**
+What should have happened instead.
+
+**Mac details**
+- macOS version: (e.g., 15.3.1)
+- Chip: (e.g., M4 Pro)
+- apple-a-day version: (`aad --version`)
+- Python version: (`python3 --version`)
+
+**Output**
+Paste the relevant `aad checkup` output or error.

apple_a_day-0.2.0/.github/ISSUE_TEMPLATE/new_check.md

@@ -0,0 +1,17 @@
+---
+name: New check request
+about: Suggest a health check apple-a-day should perform
+labels: enhancement
+---
+
+**What happened on your Mac**
+Describe the issue you encountered.
+
+**How you diagnosed it manually**
+What commands did you run? What did you find?
+
+**What apple-a-day should have told you**
+What finding summary, severity, and fix would have saved you time?
+
+**Native macOS tools involved**
+List any macOS commands or tools relevant to this check.

apple_a_day-0.2.0/.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,11 @@
+## What does this PR do?
+
+<!-- Brief description -->
+
+## Checklist
+
+- [ ] Tested on my Mac (`aad checkup` runs without errors)
+- [ ] New check follows the pattern in CONTRIBUTING.md
+- [ ] Every finding has a plain-english `summary` and a `fix`
+- [ ] No system state is modified (read-only checks only)
+- [ ] macOS-native tools only (no cross-platform abstractions)

apple_a_day-0.2.0/.github/workflows/ci.yml

@@ -0,0 +1,37 @@
+name: CI
+
+on:
+  push:
+    branches: [main]
+  pull_request:
+    branches: [main]
+
+jobs:
+  lint:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+      - run: pip install ruff
+      - run: ruff check .
+      - run: ruff format --check .
+
+  test:
+    strategy:
+      matrix:
+        os: [macos-13, macos-14, macos-15]
+        python-version: ["3.11", "3.12", "3.13"]
+    runs-on: ${{ matrix.os }}
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+      - run: pip install -e ".[dev]"
+      - run: pytest -v
+      - name: Smoke test
+        run: |
+          aad --version
+          aad checkup --json | python3 -c "import sys,json; d=json.load(sys.stdin); print(f'{len(d[\"findings\"])} findings')"

apple_a_day-0.2.0/.gitignore

@@ -0,0 +1,17 @@
+__pycache__/
+*.py[cod]
+*$py.class
+*.egg-info/
+dist/
+build/
+*.egg
+.eggs/
+*.so
+.venv/
+venv/
+env/
+.env
+.pytest_cache/
+.ruff_cache/
+.mypy_cache/
+*.DS_Store

apple_a_day-0.2.0/.ike/completed/TASK-0001 - drop-click-rewrite-cli-with-argparse.md

@@ -0,0 +1,20 @@
+---
+id: TASK-0001
+title: 'Drop click, rewrite CLI with argparse'
+status: Done
+created: '2026-03-22'
+priority: high
+milestone: MS-0001
+tags:
+  - zero-dep
+  - adr-002
+definition-of-done:
+  - aad checkup works with all existing flags
+  - click removed from pyproject.toml dependencies
+  - all 17 tests pass
+visionlog_goal_id: GOAL-007
+updated: '2026-03-22'
+---
+Replace click with stdlib argparse. Same flags: --json, --min-severity, --check, --no-parallel, --version. ADR-002 mandate.
+
+**Completion notes:** Replaced click with stdlib argparse. All flags preserved: --json, --min-severity, --check/-c, --no-parallel, --version. Added --fields. 17 tests pass.

apple_a_day-0.2.0/.ike/completed/TASK-0002 - write-ansi-formatter-make-rich-optional.md

@@ -0,0 +1,21 @@
+---
+id: TASK-0002
+title: 'Write ANSI formatter, make rich optional'
+status: Done
+created: '2026-03-22'
+priority: high
+milestone: MS-0001
+tags:
+  - zero-dep
+  - adr-002
+definition-of-done:
+  - aad checkup renders colored output without rich installed
+  - 'pip install apple-a-day[rich] enables rich tables'
+  - rich removed from core dependencies
+  - zero entries in dependencies list
+visionlog_goal_id: GOAL-007
+updated: '2026-03-22'
+---
+Write a thin ANSI escape code + Unicode box-drawing formatter as default renderer. Move rich to optional extra: pip install apple-a-day[rich]. Auto-detect: if rich is importable, use it; otherwise ANSI.
+
+**Completion notes:** format_ansi.py as default, format_rich.py auto-detected. pyproject.toml: dependencies = [], rich is optional [rich] extra. ANSI output verified independently.

apple_a_day-0.2.0/.ike/completed/TASK-0003 - add-runtime-schema-introspection-aad-schema.md

@@ -0,0 +1,20 @@
+---
+id: TASK-0003
+title: 'Add runtime schema introspection: aad schema'
+status: Done
+created: '2026-03-22'
+priority: high
+milestone: MS-0001
+tags:
+  - agent-native
+  - adr-001
+definition-of-done:
+  - aad schema outputs valid JSON describing all 7 checks
+  - each check lists its possible finding fields
+  - agents can discover capabilities without reading README
+visionlog_goal_id: GOAL-006
+updated: '2026-03-22'
+---
+Add `aad schema` command that outputs JSON schema of all checks: name, description, output fields, severity levels. Agents use this to understand what apple-a-day can do at runtime without consuming docs tokens.
+
+**Completion notes:** schema.py with get_schema(). Exposed via `aad schema`. Returns checks, output schema, usage examples, invariants.

apple_a_day-0.2.0/.ike/completed/TASK-0004 - write-skill-md-for-agent-discoverability.md

@@ -0,0 +1,21 @@
+---
+id: TASK-0004
+title: Write SKILL.md for agent discoverability
+status: Done
+created: '2026-03-22'
+priority: medium
+milestone: MS-0001
+tags:
+  - agent-native
+  - adr-001
+definition-of-done:
+  - SKILL.md exists at package root
+  - describes all commands and their flags
+  - lists operational invariants agents should know
+  - follows CLI-Anything/Poehnelt patterns from research
+visionlog_goal_id: GOAL-006
+updated: '2026-03-22'
+---
+Create SKILL.md following CLI-Anything pattern. Structured markdown that agents read to discover: what apple-a-day does, available commands, output format, invariants (read-only, never requires sudo for basic checks, always returns JSON with --json).
+
+**Completion notes:** SKILL.md at package root. Commands, checks, output format, invariants. Agent-discoverable.

apple_a_day-0.2.0/.ike/completed/TASK-0005 - add-fields-flag-for-context-window-discipline.md

@@ -0,0 +1,20 @@
+---
+id: TASK-0005
+title: Add --fields flag for context window discipline
+status: Done
+created: '2026-03-22'
+priority: medium
+milestone: MS-0001
+tags:
+  - agent-native
+  - adr-001
+definition-of-done:
+  - '--fields severity,summary returns only those fields in JSON'
+  - works with --json flag
+  - invalid field names return a clear error
+visionlog_goal_id: GOAL-006
+updated: '2026-03-22'
+---
+Add --fields flag to limit which fields are returned in JSON output (e.g. --fields severity,summary,fix). Agents use this to keep responses small and protect their context window.
+
+**Completion notes:** --fields flag filters JSON output fields. Tested: --fields severity,summary,fix correctly strips check and details.

apple_a_day-0.2.0/.ike/completed/TASK-0006 - machine-parseable-errors-structured-error-json.md

@@ -0,0 +1,22 @@
+---
+id: TASK-0006
+title: Machine-parseable errors (structured error JSON)
+status: Done
+created: '2026-03-22'
+priority: medium
+milestone: MS-0001
+tags:
+  - agent-native
+  - adr-001
+definition-of-done:
+  - 'errors return JSON with error_code, message, suggestion fields'
+  - no Python tracebacks in --json mode
+  - >-
+    agents can distinguish between 'check found problems' and 'check itself
+    failed'
+visionlog_goal_id: GOAL-006
+updated: '2026-03-22'
+---
+When a check fails to run (permission denied, tool not found, timeout), return structured JSON error with error_code, message, and suggestion — not a Python traceback or prose string. Agents need to parse errors programmatically.
+
+**Completion notes:** CheckError model with error_code, message, suggestion. Error codes: TIMEOUT, PERMISSION_DENIED, TOOL_NOT_FOUND, PARSE_ERROR, UNKNOWN_ERROR. Runner produces structured errors instead of silent swallows. JSON output includes "errors" array only when errors exist. No tracebacks in --json mode.

apple_a_day-0.2.0/.ike/completed/TASK-0007 - security-check-module.md

@@ -0,0 +1,21 @@
+---
+id: TASK-0007
+title: Security check module
+status: Done
+created: '2026-03-22'
+priority: high
+milestone: MS-0002
+tags:
+  - new-check
+  - research-0006
+definition-of-done:
+  - 'check_security() returns findings for SIP, Gatekeeper, XProtect, FileVault'
+  - SIP disabled = CRITICAL finding
+  - FileVault off = WARNING finding
+  - registered in ALL_CHECKS
+visionlog_goal_id: GOAL-008
+updated: '2026-03-22'
+---
+New check module: csrutil (SIP status), spctl (Gatekeeper), XProtect definition freshness, FileVault encryption status. From landscape research finding 0006.
+
+**Completion notes:** Security module: SIP (csrutil), Gatekeeper (spctl), FileVault (fdesetup), XProtect freshness (system_profiler install history). All four green on test Mac.

apple_a_day-0.2.0/.ike/completed/TASK-0008 - network-check-module.md

@@ -0,0 +1,20 @@
+---
+id: TASK-0008
+title: Network check module
+status: Done
+created: '2026-03-22'
+priority: high
+milestone: MS-0002
+tags:
+  - new-check
+  - research-0005
+definition-of-done:
+  - 'check_network() returns findings for throughput, latency, Wi-Fi signal'
+  - 'poor signal = WARNING, very poor = CRITICAL'
+  - registered in ALL_CHECKS
+visionlog_goal_id: GOAL-009
+updated: '2026-03-22'
+---
+New check module: networkQuality (throughput/latency), airport command (Wi-Fi signal/noise/channel). From landscape research finding 0005.
+
+**Completion notes:** Network module: Wi-Fi signal via system_profiler SPAirPortDataType (replaced deprecated airport command). SSID, RSSI, noise, SNR, channel, PHY mode. Speed test via networkQuality -s -c (JSON output). Throughput + responsiveness with Apple's RPM scale. 21 tests pass.

apple_a_day-0.2.0/.ike/documents/DOC-0001 - session-log-bootstrap-trilogy-proof-2026-03-22.md

@@ -0,0 +1,104 @@
+---
+id: DOC-0001
+title: 'Session Log: Bootstrap + Trilogy Proof — 2026-03-22'
+created: '2026-03-22'
+tags:
+  - devlog
+  - session
+  - trilogy
+---
+# Session Log: apple-a-day Bootstrap + Trilogy Proof
+
+**Date:** 2026-03-22
+**Pilot:** Daniel + Claude Opus 4.6
+
+## Origin
+
+Daniel's Mac was crashing. Investigation revealed:
+- Facebook's `watchman` crash-looping 611 times/day (broken `libboost_system.dylib` after brew upgrade)
+- `KeepAlive` launchd plist respawning it endlessly
+- 9 kernel panics in 7 days (watchdog timeouts)
+- 15.6 GB swap, 3 crash-looping services total
+
+20 minutes of manual forensics to diagnose. apple-a-day was born to make that take seconds.
+
+## What Was Built
+
+**v0.1 CLI** — 7 check modules, all working:
+- Crash loops, kernel panics, dylib health, memory pressure, disk health, launch agents, homebrew
+- Parallel execution via ThreadPoolExecutor
+- Rich terminal output + JSON mode
+- Severity filtering, selective checks, Mac info header, timing
+
+**FOSS scaffolding:**
+- MIT license, CONTRIBUTING.md, SECURITY.md, CODE_OF_CONDUCT.md
+- GitHub issue templates (bug report, new check request), PR template
+- CHANGELOG.md, GitHub Actions CI (macOS 13/14/15 × Python 3.11-3.13)
+- 17 tests passing
+
+## The Trilogy in Action
+
+### research.md → earned the decisions
+
+Three research subprojects, 22 findings total:
+
+1. **landscape** (7 findings) — 25 existing tools cataloged. Identified smartctl, networkQuality, csrutil as key gaps. Determined which tools to wrap vs reimplement.
+
+2. **foss-launch** (11 findings) — Verified top Python CLIs are zero-dep (yt-dlp 152k stars, youtube-dl 140k). Discovered pip vendored rich rather than depending on it. Found Click has a track record of breaking downstream. Mapped launch channels (HN, Reddit, awesome-lists) with specific tactics.
+
+3. **agent-operated** (5 findings) — Discovered "agent-native" is an established term (CLI-Anything/HKU). Found Poehnelt's 7 principles for agent CLIs. Mapped Agentic SRE detect→diagnose→act pattern. Confirmed 2026 industry consensus: agents are becoming the primary software consumer.
+
+### visionlog → recorded the contracts
+
+- **Vision** reframed: "Your Mac's immune system, operated by your AI agent."
+- **ADR-001:** apple-a-day is agent-native, not human-first (backed by agent-operated research)
+- **ADR-002:** Zero runtime dependencies (backed by foss-launch research)
+- **9 goals** (GOAL-001 complete, GOAL-006/007 available, rest locked)
+- **3 guardrails:** Mac-only, read-only by default, plain english always actionable
+
+### ike.md → executes within those contracts
+
+11 tasks across 2 milestones:
+
+**MS-0001 v0.2 Agent-Native + Zero-Dep (due 2026-04-05):**
+- TASK-0001: Drop click → argparse (GOAL-007, ADR-002)
+- TASK-0002: ANSI formatter, rich optional (GOAL-007, ADR-002)
+- TASK-0003: aad schema introspection (GOAL-006, ADR-001)
+- TASK-0004: SKILL.md (GOAL-006, ADR-001)
+- TASK-0005: --fields flag (GOAL-006, ADR-001)
+- TASK-0006: Structured errors (GOAL-006, ADR-001)
+
+**MS-0002 v0.3 New Checks + PyPI (due 2026-04-19):**
+- TASK-0007: Security module (GOAL-008)
+- TASK-0008: Network module (GOAL-009)
+- TASK-0009: PyPI release (GOAL-005)
+- TASK-0010: Homebrew tap (GOAL-005)
+- TASK-0011: Demo GIF
+
+## The Traceability Chain
+
+```
+research.md finding → visionlog ADR → visionlog GOAL → ike.md TASK
+```
+
+Every task traces to a goal. Every goal traces to a decision. Every decision traces to evidence. Nothing was done on vibes.
+
+## Key Insight
+
+The trilogy's value is not bureaucracy — it's **preventing drift**. In a single session the project went from "CLI for humans" to "agent-native library" based on the founder's clarification. Without the trilogy:
+- The research would be lost in chat history
+- The architectural pivot would be an undocumented gut call
+- Future agents/pilots would rebuild from scratch
+
+With the trilogy, any agent can cold-start on apple-a-day, read visionlog, and know exactly what was decided, why, and what to do next.
+
+## Stats
+
+- Repo: https://github.com/eidos-agi/apple-a-day
+- Commits: 5
+- Files: ~50
+- Lines of code: ~1200
+- Research findings: 22
+- Visionlog items: 2 ADRs, 9 goals, 3 guardrails
+- ike tasks: 11 across 2 milestones
+- Time: 1 session

apple_a_day-0.2.0/.ike/ike.json

@@ -0,0 +1,7 @@
+{
+  "id": "91197e57-081c-473f-ad83-8835f6ce7b57",
+  "version": "0.1.0",
+  "project": "apple-a-day",
+  "created": "2026-03-22",
+  "ike_path": ".ike"
+}

apple_a_day-0.2.0/.ike/milestones/MS-0001 - v0-2-agent-native-zero-dep.md

@@ -0,0 +1,8 @@
+---
+id: MS-0001
+title: 'v0.2: Agent-Native + Zero-Dep'
+status: closed
+created: '2026-03-22'
+due: '2026-04-05'
+---
+Rewrite core to zero dependencies, add agent-native affordances. This is the milestone that transforms apple-a-day from a CLI demo into an agent-native library.

apple_a_day-0.2.0/.ike/milestones/MS-0002 - v0-3-new-check-modules-pypi.md

@@ -0,0 +1,8 @@
+---
+id: MS-0002
+title: 'v0.3: New Check Modules + PyPI'
+status: open
+created: '2026-03-22'
+due: '2026-04-19'
+---
+Add security and network check modules, publish to PyPI, create Homebrew tap.

apple_a_day-0.2.0/.ike/tasks/TASK-0009 - publish-to-pypi.md

@@ -0,0 +1,16 @@
+---
+id: TASK-0009
+title: Publish to PyPI
+status: To Do
+created: '2026-03-22'
+priority: high
+milestone: MS-0002
+tags:
+  - release
+definition-of-done:
+  - pip install apple-a-day installs from PyPI
+  - aad --version shows correct version
+  - PyPI page has description and metadata
+visionlog_goal_id: GOAL-005
+---
+Publish apple-a-day to PyPI. pip install apple-a-day works. Claim the package name.

apple_a_day-0.2.0/.ike/tasks/TASK-0010 - create-homebrew-tap.md

@@ -0,0 +1,17 @@
+---
+id: TASK-0010
+title: Create Homebrew tap
+status: To Do
+created: '2026-03-22'
+priority: medium
+milestone: MS-0002
+tags:
+  - release
+  - distribution
+definition-of-done:
+  - eidos-agi/homebrew-tap repo exists
+  - brew install eidos-agi/tap/apple-a-day works
+  - formula bundles Python deps in virtualenv
+visionlog_goal_id: GOAL-005
+---
+Create eidos-agi/homebrew-tap repo with formula for apple-a-day. brew tap eidos-agi/tap && brew install apple-a-day works.

apple_a_day-0.2.0/.ike/tasks/TASK-0011 - record-demo-gif-with-asciinema.md

@@ -0,0 +1,15 @@
+---
+id: TASK-0011
+title: Record demo GIF with asciinema
+status: To Do
+created: '2026-03-22'
+priority: medium
+milestone: MS-0002
+tags:
+  - foss-launch
+definition-of-done:
+  - docs/screenshot.png or docs/demo.gif exists
+  - embedded in README.md
+  - shows real output with real findings
+---
+Record aad checkup running with colorful output. Convert to animated SVG or GIF for README. Research finding: repos with demo GIFs get ~42% more stars.

apple_a_day-0.2.0/.research/research.json

@@ -0,0 +1,15 @@
+{
+  "id": "c07b9113-7534-4c76-b7f0-3fe646a26257",
+  "version": "0.1.0",
+  "projectName": "apple-a-day-landscape",
+  "created": "2026-03-22",
+  "phase": "research",
+  "transitions": [
+    {
+      "phase": "research",
+      "date": "2026-03-22"
+    }
+  ],
+  "question": "What existing open-source tools, native macOS utilities, and community patterns should apple-a-day integrate, wrap, or learn from to build the most comprehensive Mac health toolkit?",
+  "context": "apple-a-day is a new Mac-native health toolkit in the eidos-agi ecosystem (repo: eidos-agi/apple-a-day). It already has a working v0.1 CLI (`aad checkup`) with 7 check modules: crash loops, kernel panics, dylib health, memory pressure, disk health, launch agents, and homebrew.\n\nThe motivation came from a real incident: a broken Homebrew dependency (missing libboost_system.dylib) caused Facebook's watchman to crash-loop 611 times in a day via a KeepAlive launchd plist, likely triggering kernel panics (9 in 7 days) via watchdog timeouts. The user spent 20 minutes manually diagnosing what the tool should have caught automatically.\n\nThe project has 3 guardrails: Mac-only (no cross-platform), read-only by default (fixes require opt-in), and plain-english always-actionable output.\n\nThis research catalogs 25 open-source tools and native macOS utilities across 7 categories (system monitoring, native diagnostics, disk/storage, CPU/GPU/thermal, network, launchd/crash, security) to determine what apple-a-day should wrap, invoke, mine for patterns, or reimplement. The goal is to avoid reinventing wheels while building a unified interface.\n\nKey decision: should v1 prioritize crash/panic/launchd issues (the acute pain that triggered the project) or general \"is my Mac slow/overheating/out of disk\" complaints? Or both?\n\nTarget users: Eidos AGI developers first, Mac power users and sysadmins second, general Mac users third."
+}

apple_a_day-0.2.0/.visionlog/adr/ADR-001-apple-a-day-is-agent-native-not-human-first.md

@@ -0,0 +1,35 @@
+---
+id: "ADR-001"
+type: "decision"
+title: "apple-a-day is agent-native, not human-first"
+status: "accepted"
+date: "2026-03-22"
+relates_to: ["GOAL-002", "GOAL-003", "GUARD-002"]
+source_research_id: "117c7991-5983-4dee-bc64-ba50f3021e11"
+---
+
+## Context
+
+apple-a-day was initially designed as a CLI tool for humans (`aad checkup` with Rich tables). During the first session, the founder clarified that Eidos AGI builds software that AIs run, not humans. Research into agent-native software design (CLI-Anything/HKU, Poehnelt's agent CLI manifesto, Agentic SRE patterns) confirmed this is an established and growing paradigm.
+
+## Decision
+
+apple-a-day is a **diagnostic library first**, with the AI agent as its primary consumer. Architecture is library → CLI → MCP, not CLI with MCP bolted on.
+
+Key implications:
+- Structured JSON output is the contract, rich tables are opt-in
+- Zero runtime dependencies (agents run in lean environments)
+- SKILL.md for agent discoverability
+- Runtime schema introspection (`aad schema`)
+- No interactive input — everything via flags/args
+- Checks are autonomous, fixes are policy-gated
+- The agent is not a trusted operator
+
+## Consequences
+
+- GOAL-002 (MCP server) becomes a thin wrapper, not a separate product
+- CLI output priorities shift: JSON first, tables second
+- Need to add: `--fields`, `aad schema`, SKILL.md, machine-parseable errors
+- Future `aad fix` must implement Agentic SRE trust model (least-privilege, audit trail, human approval by default)
+- Python is the right language — agents import it or shell out
+- Rich becomes optional dependency, not core

apple_a_day-0.2.0/.visionlog/adr/ADR-002-zero-runtime-dependencies-rich-and-click-are-optio.md

@@ -0,0 +1,32 @@
+---
+id: "ADR-002"
+type: "decision"
+title: "Zero runtime dependencies — rich and click are optional"
+status: "accepted"
+date: "2026-03-22"
+source_research_id: "bad85d8e-be67-4e9e-bb3a-c25eb71bd311"
+---
+
+## Context
+
+Research (foss-launch findings 0007-0011) showed:
+- Top Python CLIs are zero-dep: yt-dlp (152k stars), youtube-dl (140k), tqdm (31k)
+- pip vendored rich rather than depending on it
+- Click has a track record of breaking downstream consumers (spaCy, miiocli, others)
+- A Mac health tool that diagnoses broken dependencies should not itself have dependency problems
+
+## Decision
+
+Zero runtime dependencies. `dependencies = []` in pyproject.toml.
+- Drop click → use argparse (stdlib)
+- Move rich to optional: `pip install apple-a-day[rich]`
+- Default output uses ANSI escape codes + Unicode box-drawing
+- Follow the yt-dlp pattern: core works with nothing but stdlib Python
+
+## Consequences
+
+- Need to rewrite CLI from click to argparse (~30 lines)
+- Need to write a thin ANSI formatter as default renderer
+- Rich tables available via `apple-a-day[rich]` for humans who want them
+- Install is instant, no transitive deps, no supply chain risk
+- Marketing angle: "zero-dependency Mac health tool that diagnoses dependency problems"

apple_a_day-0.2.0/.visionlog/config.yaml

@@ -0,0 +1,3 @@
+id: "1b107e77-416b-4f01-b615-c9ce19cf784d"
+project: "apple-a-day"
+created: "2026-03-22"

apple_a_day-0.2.0/.visionlog/goals/GOAL-001-v0-1-cli-checkup-works-end-to-end.md

@@ -0,0 +1,11 @@
+---
+id: "GOAL-001"
+type: "goal"
+title: "v0.1: CLI checkup works end-to-end"
+status: "complete"
+date: "2026-03-22"
+depends_on: []
+unlocks: []
+---
+
+All 7 check modules run, produce findings, and render in the terminal via `aad checkup`. JSON output works. Installable via pip.

apple_a_day-0.2.0/.visionlog/goals/GOAL-002-mcp-server-claude-sessions-can-query-mac-health.md

@@ -0,0 +1,11 @@
+---
+id: "GOAL-002"
+type: "goal"
+title: "MCP server: Claude sessions can query Mac health"
+status: "locked"
+date: "2026-03-22"
+depends_on: []
+unlocks: []
+---
+
+Thin MCP wrapper over the diagnostic library (per ADR-001). Expose each check as an MCP tool: `mac_checkup`, `mac_crash_loops`, `mac_kernel_panics`, etc. Returns structured Finding objects. This is one of three surfaces (import, CLI, MCP), not the primary interface.

apple_a_day-0.2.0/.visionlog/goals/GOAL-003-auto-fix-opt-in-remediation-for-common-issues.md

@@ -0,0 +1,11 @@
+---
+id: "GOAL-003"
+type: "goal"
+title: "Auto-fix: opt-in remediation for common issues"
+status: "locked"
+date: "2026-03-22"
+depends_on: []
+unlocks: []
+---
+
+Per ADR-001 and Agentic SRE research: implement the "act" phase of detect→diagnose→act. Fixes are policy-gated: human approval by default, agent can request override. Every fix logged with before/after state for audit. `--dry-run` required on all mutations. The agent is not a trusted operator.

apple_a_day-0.2.0/.visionlog/goals/GOAL-004-thermal-power-monitoring-apple-silicon-.md

@@ -0,0 +1,11 @@
+---
+id: "GOAL-004"
+type: "goal"
+title: "Thermal & power monitoring (Apple Silicon)"
+status: "locked"
+date: "2026-03-22"
+depends_on: []
+unlocks: []
+---
+
+Leverage `powermetrics`, IOKit, and SMC to surface CPU/GPU temperature, throttling state, fan speed, and power draw. Apple Silicon specific.

apple_a_day-0.2.0/.visionlog/goals/GOAL-005-pypi-release.md

@@ -0,0 +1,11 @@
+---
+id: "GOAL-005"
+type: "goal"
+title: "PyPI release"
+status: "locked"
+date: "2026-03-22"
+depends_on: []
+unlocks: []
+---
+
+Publish to PyPI as `apple-a-day` so anyone can `pip install apple-a-day` and run `aad`.

apple_a_day-0.2.0/.visionlog/goals/GOAL-006-agent-native-affordances-schema-introspection-skil.md

@@ -0,0 +1,16 @@
+---
+id: "GOAL-006"
+type: "goal"
+title: "Agent-native affordances: schema introspection, SKILL.md, --fields"
+status: "available"
+date: "2026-03-22"
+depends_on: []
+unlocks: []
+---
+
+Add the agent-native features identified in research:
+- `aad schema` — runtime introspection of all checks and their output schemas
+- SKILL.md — co-located agent-discoverable capability definition
+- `--fields` flag — limit response fields to protect agent context window
+- Machine-parseable errors (error codes + structured JSON, not prose)
+- No interactive input anywhere

apple_a_day-0.2.0/.visionlog/goals/GOAL-007-zero-dep-rewrite-drop-click-rich-argparse-ansi.md

@@ -0,0 +1,11 @@
+---
+id: "GOAL-007"
+type: "goal"
+title: "Zero-dep rewrite: drop click/rich, argparse + ANSI"
+status: "available"
+date: "2026-03-22"
+depends_on: []
+unlocks: []
+---
+
+Per ADR-002: replace click with argparse, replace rich with ANSI escape codes + Unicode box-drawing. Move rich to optional dependency under `[rich]` extra. Core runs on stdlib Python only.