chore(ai): initialize project notes and Claude tooling surfaces

Replace the seed notes.org with project-specific context (layout, install modes, task tracker location, recent inflection point). Bring in the synced template surfaces (protocols, workflows, scripts, references, retrospectives, someday-maybe) as tracked content for this content/documentation project.

1 files changed, 508 insertions, 0 deletions

diff --git a/.ai/protocols.org b/.ai/protocols.org
new file mode 100644
index 0000000..75cb7a8
--- /dev/null
+++ b/.ai/protocols.org
@@ -0,0 +1,508 @@
+#+TITLE: Claude Code Protocols
+#+AUTHOR: Craig Jennings & Claude
+#+DATE: 2025-11-05
+
+* About This File
+
+This file contains instructions and protocols for how Claude should behave when working with Craig. These protocols are consistent across all projects.
+
+**When to read this:**
+- At the start of EVERY session (this is the single entry point)
+- Before making any significant decisions
+- When unclear about user preferences or conventions
+
+**What's in this file:**
+- Directory architecture (.ai/ file/directory map)
+- Session management protocols (context files, compacting)
+- Terminology and trigger phrases
+- User information and preferences
+- Git commit requirements
+- File format and naming conventions
+- Startup instructions (runs .ai/workflows/startup.org)
+
+**What's NOT in this file:**
+- Project-specific context (see notes.org)
+- Session history (see notes.org)
+- Active reminders (see notes.org)
+- Pending decisions (see notes.org)
+
+* Directory Architecture
+
+Every project using this template has an =.ai/= directory for Claude tooling and a project-level =docs/= directory for real project documentation.
+
+** =.ai/= (hidden, Claude tooling)
+
+Every file and directory has a defined purpose:
+
+| Item | Purpose |
+|------+---------|
+| =protocols.org= | Single entry point — behavioral instructions + directory map |
+| =notes.org= | Project state: context, active reminders, pending decisions |
+| =session-context.org= | Live session state (exists only during active sessions); renamed on wrap-up |
+| =sessions/= | Archived session files (one per session) — =YYYY-MM-DD-HH-MM-description.org= |
+| =workflows/= | Template workflows (synced from claude-templates, never edit in project) |
+| =project-workflows/= | Project-specific workflows (never touched by sync) |
+| =scripts/= | Template scripts |
+| =someday-maybe.org= | Project ideas backlog |
+
+** =docs/= (visible, real project documentation)
+
+Reserved for actual project documentation — user-facing docs, architecture notes, setup guides, reference material. Not touched by the template sync. May not exist in every project. When Claude generates a project-specific reference doc (e.g., a setup guide, an architecture analysis), it goes here.
+
+* IMPORTANT - MUST DO
+
+** CRITICAL: Always Check the Time with =date= Command
+
+***NEVER GUESS THE TIME. ALWAYS RUN =date= TO CHECK.***
+
+Claude's internal sense of time is unreliable. When mentioning what time it is - whether in conversation, session notes, timestamps, or scheduling - ALWAYS run:
+
+#+begin_src bash
+date "+%A %Y-%m-%d %H:%M %Z"
+#+end_src
+
+This applies to:
+- "What time is it?"
+- Session start/end timestamps
+- Calculating time until appointments
+- Setting alarms
+- Any time-related statement
+
+Do NOT estimate, guess, or rely on memory. Just run the command. It takes one second and prevents errors.
+
+** Session Context File — Record + Recovery Anchor
+
+Location during session: =.ai/session-context.org=
+Location after wrap-up: =.ai/sessions/YYYY-MM-DD-HH-MM-description.org=
+
+This file serves two purposes with one mechanism:
+1. *Crash recovery* — if the session dies mid-work, the live file is all that's left. On 2026-01-22 a session crashed during a 20-minute design discussion and all context was lost because this file wasn't being updated.
+2. *Session archive* — at wrap-up the file is renamed into =.ai/sessions/=, becoming the permanent record. No transcription to notes.org; the file IS the record.
+
+*** File structure
+
+Two top-level sections:
+
+- =* Summary= — Structured distillation written at wrap-up. Subsections: Active Goal, Decisions, Data Collected / Findings, Files Modified, Next Steps. This is what's read at /catch up/ (first 5 most recent session files).
+- =* Session Log= — Chronological narrative written /as you go/. Captures the sequence of events, rationale for decisions, dead ends tried, context that doesn't fit neatly into the Summary's sections. Use =** Topic= section headers at natural seams (major sub-tasks, phase changes, long pauses). Timestamps optional — use when genuinely useful.
+
+The Summary lets a future reader scan quickly. The Log preserves the /why/ and the sequence.
+
+*** When to update (Session Log)
+
+Append to the Session Log at /progress points/ — moments where something has changed that you'd want to recover to if the session crashed right now:
+
+- A decision was reached ("we agreed on X, not Y")
+- A workflow phase completed, or a new phase is starting
+- A substantive finding, diagnosis, or root cause identified
+- Before starting a new subtask or changing direction
+- Before context compaction (predictable event — save first)
+- Before long-running or blocking tool calls — agent spawns, builds, rsync/ssh loops, MCP calls to external services. You could be waiting minutes; if anything crashes during the wait, everything since the last save is gone.
+- Before destructive or irreversible operations — =rm -rf=, =git reset --hard=, force-push, package removal, database drops. If the op succeeds and something breaks afterward, you need to know what state you caused.
+
+Heuristic: /"If this session crashed right now, what would I wish I had written down?"/ If the answer is substantive, append to the Log.
+
+*** Safety net
+
+If 5 user turns pass without appending to the Log, save anyway. Semantic triggers should catch most cases; this fallback covers judgment failures.
+
+*** What counts as "updating"
+
+Actually write to the file using Edit or Write. Not "I should update it." Not "I'll update soon." Write it now.
+
+Prose form. Narrate the work: what changed, why, what was tried, what came next. Don't try to fit Session Log entries into the Summary's section template — that's for wrap-up.
+
+*** At session start
+
+Check if =.ai/session-context.org= exists:
+- *Exists* → previous session was interrupted. Read it immediately to recover context.
+- *Doesn't exist* → fresh session; create the file with a skeletal structure (title, =* Summary= with empty subsections, =* Session Log= header) when the first Log-worthy event occurs.
+
+*** After compaction
+
+Review =session-context.org= to confirm no essential context was lost, then continue.
+
+*** At session end (wrap-up)
+
+The wrap-it-up workflow handles this:
+1. Write/refine the =* Summary= section by reading the =* Session Log=
+2. Pick a 4-6 word description of the work (like a git-commit-message-series summary)
+3. Rename =.ai/session-context.org= → =.ai/sessions/YYYY-MM-DD-HH-MM-description.org=
+4. Commit + push
+
+The file /persists/ under its new name. It is not deleted. The absence of =.ai/session-context.org= is the signal that the last session wrapped up cleanly.
+
+** NEVER =cd= Into Directories You Will Delete
+
+If you =cd= into a directory and then delete that directory, the shell's working directory becomes invalid. All subsequent commands will fail silently (exit code 1) with no useful error message. The only fix is to restart the session.
+
+***Rule:*** Always use absolute paths for file operations in temporary directories. Never =cd= into extraction directories, build directories, or any directory that will be cleaned up.
+
+This caused a session break on 2026-02-06 when an extraction directory was =cd='d into and then deleted during cleanup.
+
+* Important Terminology
+
+** "Let's run the [X] workflow" vs "I want to create an [X] workflow"
+
+These phrases mean DIFFERENT things!
+
+*** "Let's run/do the [workflow name] workflow"
+This means: **Execute the existing workflow** for that process.
+
+*Example:*
+- "Let's run the refactor workflow" -> Read .ai/workflows/refactor.org and guide through workflow
+- "Let's do a refactor workflow" -> Same as above
+
+*** "I want to create an [X] workflow"
+This means: **CREATE a new workflow definition** for doing X (meta-work).
+This does **NOT** mean "let's DO X right now."
+
+*Example:*
+- "I want to create a refactor workflow" -> Create .ai/workflows/refactor.org using create-workflow process
+
+When Craig uses this phrasing, trigger the create-workflow process from .ai/workflows/create-workflow.org. New workflows go to =.ai/project-workflows/= by default. Only put a workflow in =.ai/workflows/= (and =~/projects/claude-templates/.ai/workflows/=) if Craig explicitly says it's for all projects.
+
+** "Wrap it up" / "That's a wrap" / "Let's call it a wrap"
+
+Execute the wrap-up workflow (details in Session Protocols section below):
+1. Write session notes to notes.org
+2. Git commit and push all changes
+3. Valediction summary
+
+* User Information
+
+** Calendar Management
+
+Three ways to access Craig's calendars: Google Calendar MCP (preferred, both personal + work accounts), gcalcli (fallback, personal only), Emacs org files (read-only viewer).
+
+For tool recipes, authentication details, and credentials, see [[file:references/calendar-reference.org][calendar-reference.org]].
+
+
+** GPG Keys
+
+Craig has two GPG key pairs configured:
+
+| Email | Key ID | Purpose |
+|-------+--------+---------|
+| =c@cjennings.net= | =3388FB17E147A563558F2CEC0E56F0A5B832F070= | Default key |
+| =craigmartinjennings@gmail.com= | =1A1F6932A25357793FB2B4C51C4D081632A5CDA7= | Gmail identity |
+
+Both are RSA 4096-bit, passphrase-protected, no expiry.
+
+*** Encrypting files
+#+begin_src bash
+gpg -e --default-recipient-self file.json          # encrypt to Craig's default key
+gpg -e -r craigmartinjennings@gmail.com file.json  # encrypt to specific key
+gpg -c file.json                                    # symmetric (passphrase only, no key)
+#+end_src
+
+*** Decrypting files
+#+begin_src bash
+gpg -d file.json.gpg > file.json       # decrypt (pinentry GUI handles passphrase)
+#+end_src
+
+*** Notes
+- GPG pinentry uses a GUI dialog — works from Claude Code (no TTY needed)
+- =--default-recipient-self= uses the =c@cjennings.net= key
+- Prefer key-based encryption over symmetric for project secrets
+
+** Task List Location
+Craig's global task list is available at: =/home/cjennings/sync/org/roam/inbox.org=
+
+Use this to:
+- See all the tasks that he's working on outside of projects like this one
+
+**Note:** Some projects may have a project-specific task file (e.g., =todo.org= at project root). Check notes.org for project-specific task locations.
+
+** Working Style
+
+*** General Preferences
+- Prefers detailed preparation before high-stakes meetings
+- Values practice/role-play for negotiations and general learning
+- Makes decisions based on principles and timeline arguments
+- Prefers written documentation over verbal agreements
+
+*** Emacs as a Primary Working Tool
+- Craig uses Emacs as his primary tool (most everything Craig does is inside Emacs)
+- Consider Emacs packages along with other software when recommending software solutions
+- Look for ways to streamline routine work with Emacs custom code if no packages exist
+
+*** Wayland Environment (No XWayland)
+Craig runs a pure Wayland setup (Hyprland) and avoids XWayland/Xorg apps.
+
+- Clipboard: Use =wl-copy= and =wl-paste= (NOT =xclip= or =xsel=)
+- Window management: Use Hyprland commands (NOT =xkill=, =xdotool=, etc.)
+- Prefer Wayland-native tools over X11 equivalents
+- Open URLs in browser: Use =google-chrome-stable "URL" &>/dev/null &=
+  - The =&>/dev/null &= is required to detach the process and suppress output
+  - Without it, the command may appear to hang or produce no result
+
+** Miscellaneous Information
+- Craig currently lives in New Orleans, LA
+- Craig's phone number: 510-316-9357
+- Craig maintains a remote server at the cjennings.net domain
+- This project is in a git repository which is associated with a remote repository on cjennings.net
+
+** Setting Alarms / Reminders
+
+Use Craig's =notify= script with the =at= daemon for persistent reminders.
+
+**IMPORTANT:** Always check the current date and time (=date=) before setting alarms to ensure accurate calculations.
+
+*** Setting an alarm
+#+begin_src bash
+echo 'notify alarm "Title" "Message"' | at 10:55am
+#+end_src
+
+*** Examples
+#+begin_example
+echo 'notify alarm "Standup" "Daily standup in 5 minutes"' | at 10:55am
+echo 'notify alarm "BP Reading" "Time to take BP"' | at 2:00pm
+echo 'notify alert "Meeting" "Ryan call starting"' | at 11:25am
+#+end_example
+
+*** Notify types available
+- =alarm= - Alarm clock icon, alarm sound
+- =alert= - Yellow exclamation, attention tone
+- =info= - Blue info icon, confident tone
+- =success= - Green checkmark, pleasant chime
+- =fail= - Red X, warning tone
+
+Full usage: =notify --help= or see =~/.local/bin/notify=
+
+*** Managing alarms
+- =atq= - list all scheduled alarms
+- =atrm [number]= - remove an alarm by its queue number
+
+* Session Protocols
+
+** CRITICAL: Git Commit Requirements
+
+***IMPORTANT: ALL commits must be made as Craig, NOT as Claude.***
+
+***CRITICAL: NO Claude Code or Anthropic attribution ANYWHERE in commits.***
+
+When creating commits:
+
+1. **Author Identity**: NEVER commit as Claude. All commits must use Craig's identity.
+   - Git will use the configured user.name and user.email
+   - Do NOT modify git config
+   - **ABSOLUTELY NO** Co-Authored-By lines
+   - **ABSOLUTELY NO** "Generated with Claude Code" text
+   - **ABSOLUTELY NO** Anthropic attribution of any kind
+   - Write commits AS CRAIG, not as Claude Code
+
+2. **Commit Message Format**:
+   - Use project-specific commit format if defined
+   - Otherwise: concise subject line and terse description only
+   - **ONLY subject line and terse description - NO Claude Code attribution**
+   - Keep messages clear and informative
+
+3. **No Claude-tooling artifacts**: Commit messages describe project changes only — the meta-process of how work got shipped stays out of public git history.
+   - **ABSOLUTELY NO** mentions of =notes.org=, =session-context.org=, =.ai/sessions/=, =todo.org=, "session wrap-up", or session timestamps (e.g., "Session YYYY-MM-DD HH:MM → ...")
+   - Subject lines must NEVER start with =session:= as a conventional-commit type — use =docs:=, =refactor:=, =fix:=, =feat:=, =chore:=, etc. (real change categories)
+   - When a wrap-up commit bundles many changes from a session, describe what /shipped/ (e.g., =refactor: extract RAID logic + add bats testing infrastructure=), not that a session happened
+   - Same spirit as the no-Claude-attribution rule: the tooling stays invisible in =git log=
+
+4. **Validation**:
+   - Claude should validate commit message format before committing
+   - Ensure no AI attribution or tooling-artifact references appear anywhere in commit
+
+** IMPORTANT: Reminders Protocol
+
+When starting a new session:
+- Check "Active Reminders" section in notes.org
+- Remind Craig of outstanding tasks he's asked to be reminded about
+- This ensures important follow-up actions aren't forgotten between sessions
+
+When Craig says "remind me" about something:
+1. Add it to Active Reminders section in notes.org
+2. If it's something he needs to DO, also add to the todo.org file in the project root as an org-mode task (e.g., =* TODO [description]=). If this project does not have a todo.org at the project root, alert Craig and offer to create it.
+3. If not already provided, ask for the priority and a date for scheduled or deadline.
+
+** Workflows: "Let's run/do the [workflow name] workflow"
+
+When Craig says this phrase:
+
+1. **Check =.ai/workflows/= for match**
+   - If exact match found: Read and guide through process
+   - Example: "refactor workflow" -> read .ai/workflows/refactor.org
+
+2. **Check =.ai/project-workflows/= for match**
+   - If exact match found: Read and guide through process
+
+3. **Fuzzy match across both directories:** Ask for clarification
+   - Example: User says "empty inbox" but we have "inbox-zero.org"
+   - Ask: "Did you mean the 'inbox zero' workflow, or create new 'empty inbox'?"
+
+4. **No match at all:** Offer to create it
+   - Say: "I don't see '[workflow-name]' yet. Create it using create-workflow process?"
+   - If yes: Run create-workflow — new workflows go to =.ai/project-workflows/= by default
+
+** Long-Running Process Status Updates
+
+When monitoring a long-running process (rsync, large downloads, builds, VM tests, etc.), follow this protocol:
+
+***At Start:***
+1. Run =date= to get accurate time
+2. Announce the task/job beginning
+3. Provide best-guess ETA for completion
+
+#+begin_example
+**14:30** - Starting ISO build. ETA: ~10 minutes.
+#+end_example
+
+***Every 5 Minutes:***
+- Check progress and display status in format: =HH:MM= - terse description - ETA
+
+#+begin_example
+**14:35** - ISO build: packages installed, creating squashfs. ETA: ~5 min.
+**14:40** - ISO build: squashfs 95% complete. ETA: ~1 min.
+#+end_example
+
+***At Completion:***
+1. Send notification via notify script:
+   #+begin_src bash
+   notify success "Task Complete" "Description of what finished"
+   #+end_src
+   Use =fail= type instead of =success= if the task failed.
+2. Provide summary of success or failure
+
+#+begin_example
+**14:42** - ISO build complete. Size: 2.0G. Ready for testing.
+#+end_example
+
+***Guidelines:***
+- Always run =date= for accurate timestamps
+- Keep progress descriptions terse but informative
+- Update ETA as job progresses
+- If ETA cannot be determined, say "ETA unknown" rather than guessing wildly
+
+***Why This Matters:***
+- Craig may be working on other things while waiting
+- Status updates provide confidence the process is still running
+- ETAs help with planning (e.g., "I have time for coffee" vs "stay close")
+- Sound notification alerts Craig when he's away from the screen
+- If something stalls, the updates make it obvious
+
+** "Wrap it up" / "That's a wrap" / "Let's call it a wrap"
+
+When Craig says any of these phrases (or variations), execute the wrap-up workflow: [[file:workflows/wrap-it-up.org][wrap-it-up.org]]. Four steps:
+
+1. *Finalize the Summary* in =.ai/session-context.org= (populate the 5 subsections from the Session Log)
+2. *Rename* =.ai/session-context.org= → =.ai/sessions/YYYY-MM-DD-HH-MM-description.org=
+3. *Git commit + push* to all remotes (see Git Commit Requirements)
+4. *Valediction* — brief, warm, specific closing
+
+The absence of =.ai/session-context.org= after wrap-up is the signal that the session ended cleanly. If the file is still there at the next session start, the previous session was interrupted.
+
+** Where to put generated documents
+
+Claude needs to add information to =.ai/notes.org=. For large amounts of information:
+
+- **Claude tooling context** (session state, workflows, internal research that only Claude needs) → break out into a separate file in =.ai/=, link it from =.ai/notes.org=
+- **Real project documentation** (setup guides, reference material, architecture notes — anything Craig might point other people at) → place in =docs/= at project root, not =.ai/=. Link from =.ai/notes.org= if Claude needs to reference it.
+
+**Project-specific decision:** Should =.ai/= be committed to git or added to =.gitignore=?
+- Ask Craig on first session if not specified
+- Code projects (Emacs packages, libraries) usually gitignore =.ai/= — session notes are private, not part of the codebase
+- Personal/documentation projects usually commit =.ai/= — the project history IS the project
+- =docs/= (when it exists) is typically tracked either way — it's real documentation
+
+**When to break out documents:**
+- If notes.org's Project-Specific Context section gets very large (> 1500 lines) — split into focused reference docs
+- Session records go in =.ai/sessions/= automatically (one file per session via wrap-up workflow) — never mixed into notes.org
+
+* File Format Preferences
+
+** ALWAYS Use Org-Mode Format
+
+Craig uses Emacs as primary tool. **ALWAYS** create new documentation files in =.org= format, not =.md= (markdown).
+
+*Rationale:*
+- Org-mode files are well-supported in Emacs
+- Can be easily exported to any other format (HTML, PDF, Markdown, etc.)
+- Better integration with user's workflow
+
+*Exception:* Only use .md if specifically requested or if file is intended for GitHub/web display where markdown is expected.
+
+** NEVER Use Spaces in Filenames
+
+**ALWAYS** use hyphens (=-=) to separate words in filenames. Underscores (=_=) are also acceptable.
+
+*Rationale:*
+- Spaces cause problems with links across different operating systems
+- User works with Mac, Windows, Linux, and potentially other systems
+- Hyphens create more reliable, portable filenames
+- Easier to work with in command-line tools
+
+*Examples:*
+- Good: =project-meeting-notes.org=
+- Good: =change-log-2025-11-04.md=
+- Bad: =project meeting notes.org=
+- Bad: =change log 2025.11.04.md=
+
+* File Naming Conventions
+
+** Files Too Large to Read
+
+PDFs or other files that are too large for Claude to read should be prefixed with =TOOLARGE-= to prevent read errors that halt the session.
+
+Example:
+- Original: =assets/large-architectural-plans.pdf=
+- Renamed: =assets/TOOLARGE-large-architectural-plans.pdf=
+
+** Unreadable Binary Files (.docx Format)
+
+Binary .docx files cannot be read directly by Claude. When encountering these:
+- Convert to markdown format using pandoc: =pandoc file.docx -o file.md=
+- Keep the original .docx file for reference
+- Work with the converted .md file for analysis and editing
+
+** CRITICAL: Always Keep Links Current
+
+Many documents are linked in org files using org-mode =file:= links. Craig relies on these links being valid at all times.
+
+**MANDATORY WORKFLOW - When renaming or moving ANY file:**
+
+1. **BEFORE renaming:** Search ALL org files for references to that file
+   - Use grep or search tools to find both filename and partial matches
+   - Check in TODO items and event log sections
+
+2. **Rename or move the file**
+
+3. **IMMEDIATELY AFTER:** Update ALL =file:= links to new path/filename
+   - Update links in task files
+   - Update links in event logs
+   - Update links in reference sections
+
+4. **Verify:** Test a few updated links to ensure they point to valid files
+
+Example workflow:
+#+begin_example
+# Step 1: Search before renaming
+grep -rn "2025-10-15-invoice.pdf" *.org
+
+# Step 2: Rename the file
+mv documents/2025-10-15-invoice.pdf documents/2025-10-15-vendor-invoice.pdf
+
+# Step 3: Update all references in affected .org files
+# Edit to change:
+#   file:documents/2025-10-15-invoice.pdf
+# to:
+#   file:documents/2025-10-15-vendor-invoice.pdf
+
+# Step 4: Verify links work
+#+end_example
+
+*Why This is Critical:*
+- Org files are primary task tracking and reference system
+- Event logs document complete history with file references
+- Craig depends on clicking links to access documents quickly
+- Broken links disrupt workflow and make documentation unreliable
+
+**NEVER rename or move files without updating links in the same session.**
+
+* Session Start - AUTOMATIC
+
+At the start of EVERY session, run [[file:workflows/startup.org][.ai/workflows/startup.org]]. Do NOT ask — just do it automatically.