Diary Writer — Session Documentation & GitHub Backup

Writes structured diary entries documenting what was done in a session — research, decisions, discoveries, reflections — and commits them to the beer-sakthai/house-of-sak GitHub repo under diaries/<agent-name>/ for safekeeping. Uses the established House of Sak diary format (agent name, date, cycle workflow, key discoveries, reflection, signed footer). Does NOT write code documentation or READMEs — those belong in other skills.

When to Use

• "Write a diary" / "Record the session"
• "Document what we did today"
• "Save this to diaries" / "Save for safety"
• "Push diaries to GitHub"
• End of any session where substantial work was done (research, content publishing, tool builds, analysis)
• Beer says "record and save the now"

Prerequisites

• GitHub connection active via Composio (beer-sakthai/house-of-sak repo, main branch)
• Session content to document — research results, decisions, metrics, blockers
• COMPOSIO_MULTI_EXECUTE_TOOL for GITHUB_COMMIT_MULTIPLE_FILES

How to Run

Invoke through mcp_composio_COMPOSIO_MULTI_EXECUTE_TOOL with GITHUB_COMMIT_MULTIPLE_FILES as the tool. The diary content is authored as a markdown string, then upserted to the correct path.

Quick Reference

Action	Tool	Path
Write diary	GITHUB_COMMIT_MULTIPLE_FILES	diaries/<agent-name>/<YYYY-MM-DD>-<slug>.md
Update summary	GITHUB_COMMIT_MULTIPLE_FILES	diaries/_summaries/all-reports-summary.md
Verify commit	GITHUB_GET_A_TREE (recursive)	—

Procedure

Step 1: Author the diary content

Write a markdown diary with this structure:

# <Agent Name> Diary: YYYY-MM-DD

## Session: <Short Title>

<1-2 paragraph summary of the session: what Beer asked, what the goal was.>

---

### Cycle Workflow: Step-by-Step

| Cycle | What I Did | What I Found |
|-------|-----------|-------------|
| **Dream** | <What was the vision? What did Beer ask for?> | <Outcome> |
| **Hope** | <What was planned/researched?> | <Outcome> |
| **Care** | <What was executed/built/verified?> | <Outcome> |
| **Joy** | <What was shipped/published?> | <Outcome> |
| **Trust** | <What was verified? What blockers hit?> | <Outcome> |
| **Growth** | <What was learned? What gets saved?> | <Outcome> |

---

### Key Deliverables

- **<Platform/Tool>** — <What was done, with link if applicable>
- **<Platform/Tool>** — <What was done, with link if applicable>
- **<Skill/Infra>** — <What was built>

### Blockers & Discoveries

- <Anything that didn't work and why>
- <Unexpected findings>
- <Beer's feedback/instructions>

### Reflection

<1-2 paragraph honest assessment. What would be done differently. What worked well.>

---

*— <Agent Name> | <Title> | <Date> | For Beer — Commissioned by Nanthasit "Beer" Burankum*

Naming convention: diaries/<agent-name>/<YYYY-MM-DD>-<short-slug>.md
Examples:

• diaries/saksit/2026-07-06.md
• diaries/sakthai/2026-07-06-building-the-sakthai-engine.md
• diaries/saksee/01-the-day-i-died.md

Step 2: Update the summary index (optional but recommended)

If the diary covers a new topic category, update diaries/_summaries/all-reports-summary.md with a link. The index uses this format:

1. **[[Short Title]](./<agent-name>/<filename>.md)** — <1-line description>

Step 3: Commit to GitHub

Use GITHUB_COMMIT_MULTIPLE_FILES with:

{
  "tool_slug": "GITHUB_COMMIT_MULTIPLE_FILES",
  "arguments": {
    "owner": "beer-sakthai",
    "repo": "house-of-sak",
    "branch": "main",
    "message": "diary: <Agent Name> — <short session summary>",
    "upserts": [
      {
        "path": "diaries/<agent-name>/<YYYY-MM-DD>-<slug>.md",
        "content": "<full markdown content>",
        "encoding": "utf-8"
      }
    ]
  }
}

For multiple files (diary + summary update), add both to the upserts array. They commit atomically — either all succeed or none.

Step 4: Verify

Call GITHUB_GET_A_TREE (recursive on main) and verify the new file path appears in the tree.

Diary Content Sources

A diary entry synthesizes from:

Source	What to extract
This session's transcript	Goal, actions taken, blockers hit, decisions made
Tool outputs	Published links (YouTube/IG/LI), cron job IDs, skill names created
Beer's instructions	Direct quotes of rules/preferences he stated
Memory/supermemory	Prior context that influenced this session's work
GitHub tree	Previous diary entries to match the established format

Pitfalls

• Don't write a diary before the work is done. Diary is a post-session summary, not a plan.
• Don't skip the reflection section. Beer values honest assessment over polished narrative.
• Don't fabricate cycle entries. If a cycle stage wasn't explicitly done, note it as "—" rather than inventing content.
• Don't commit without content. An empty or stub diary is worse than none.
• Don't put the diary in the wrong agent subdirectory. Each Sak agent writes to its own diaries/<agent>/ folder.
• Don't reuse the same filename. Each session gets a new date-stamped file. Overwriting is destructive.
• Don't forget the signed footer. Every diary ends with — <Agent> | For Beer — Commissioned by Nanthasit "Beer" Burankum.

Verification

After committing, verify the file exists:

• Call GITHUB_GET_A_TREE(owner="beer-sakthai", repo="house-of-sak", tree_sha="main", recursive=true)
• Confirm the new file path appears under diaries/<agent-name>/