diary

Diary

Overview

Produce concise, structured diary steps that capture the implementation journey (including failures) and make review and continuation straightforward.

Create and Maintain a Diary (docmgr)

Create the diary document

docmgr doc add \
  --ticket TICKET-ID \
  --doc-type reference \
  --title "Diary"

Relate files (always use absolute paths)

docmgr doc relate \
  --doc ttmp/.../reference/01-diary.md \
  --file-note "/abs/path/to/file.go:Why this file matters"

Update changelog when a step completes

docmgr changelog update --ticket TICKET-ID \
  --entry "Step N: ... (commit <hash>)" \
  --file-note "/abs/path/to/file.go:Reason"

Recommended Working Loop (keeps code ↔ docs consistent)

Use this loop for any non-trivial change:

# 1) Implement + format + test (capture failures verbatim)
gofmt -w path/to/file.go && go test ./... -count=1

# 2) Commit code (focused)
git add path/to/file.go && git commit -m "..." && git rev-parse HEAD

# 3) Update ticket bookkeeping
docmgr task check --ticket TICKET-ID --id N

# 4) Update diary + relate files + update changelog (include commit hash)
docmgr doc relate --doc ttmp/.../reference/01-diary.md \
  --file-note "/abs/path/to/file.go:What changed (commit <hash>)"
docmgr changelog update --ticket TICKET-ID \
  --entry "Step N: ... (commit <hash>)" \
  --file-note "/abs/path/to/file.go:Reason"

# 5) Commit docs changes
git add ttmp/... && git commit -m "Diary: record Step N"

Diary Step Format (strict)

When asked to “write/update the diary”, output Markdown that follows this structure:

# Diary

## Goal

Brief statement of what this diary captures.

## Step N: [Descriptive Step Name]

Write 1–2 short prose paragraphs here (required). Explain intent, what changed at a high level, and what it unlocked.

### Prompt Context

**User prompt (verbatim):** "<paste exactly what the user typed>"

**Assistant interpretation:** <your interpretation of what the user is asking for (paraphrase)>

**Inferred user intent:** <what outcome the user is trying to achieve / why they asked>

**Commit (code):** <hash> — "<message>"   # if code changed

### What I did
- Concrete actions (files changed, commands run)

### Why
- Motivation / rationale

### What worked
- Outcomes that moved the work forward

### What didn't work
- Failures/blockers with exact errors, versions, and commands

### What I learned
- Insights, corrected assumptions, patterns

### What was tricky to build
- Sharp edges: invariants, ordering constraints, normalization quirks, deadlocks, etc.

### What warrants a second pair of eyes
- Review-critical correctness/perf/concurrency risks

### What should be done in the future
- Only follow-ups implied by this step; if nothing: `N/A`

### Code review instructions
- Where to start (files + key symbols)
- How to validate (commands/tests)

### Technical details
- Reference material (snippets, schema/contracts, command examples)

Writing Rules (strict)

• Start each step with 1–2 short prose paragraphs (required).
• Always include a Prompt Context section for each step:
  • User prompt (verbatim) must be the exact user text (no paraphrasing, no normalization).
  • Assistant interpretation is your concise paraphrase of the request.
  • Inferred user intent captures the user’s underlying goal/outcome.
• Record failures immediately; keep errors/versions/commands verbatim.
• Always include these sections for any code/behavior change: What was tricky to build, What warrants a second pair of eyes, What should be done in the future.
• Do not add backwards-compatibility shims unless the ticket/spec requires it; document behavior changes and update tests/docs instead.
• Relate any file you modified and any file that materially shaped decisions; always use absolute paths in --file-note.

Reference

Load references/diary.md for the full diary rationale, patterns, and a worked example entry.