Diary Note

Diary Note

快速追记一段内容到今天的日记。适合 agent session 结束时记录经验、随手记 TIL、 或任何不想走完整日记流程的场景。

Prerequisites

| Tool | Type | Required | Install | |------|------|----------|---------| | None | — | — | No external dependencies |

When to Use

• Agent session 结束后，想把经验/教训记下来
• 随手记一个 TIL 或想法
• 工作中遇到值得记录的事，但不想启动完整日记流程

不适用：

• 完整日记写作 → 使用 diary-assistant
• 周报写作 → 使用 weekly-report

Flow

1. Locate diary file   (~0min, automatic)
2. Determine section   (~0min, automatic)
3. Append content      (~1min)
4. Optional: Anki      (~0min, ask only)

Step 1: Locate Diary File and Style Preferences

Diary configuration is shared with diary-assistant.

Check for config at ~/.config/nini-skill/diary/user-config.md. If not found, check diary-assistant's references/user-config.md as fallback. If neither exists, ask the user for their diary path.

After locating the config, also read the ## 写作风格 section if present. Layer those personal style preferences on top of the universal rules in Step 3 (formatting, punctuation, voice cues). If the section is absent, the universal rules alone are enough.

File name: {YYYY-MM-DD}.md (today's date).

If the file doesn't exist, create it with a basic heading:

# {YYYY-MM-DD}

Step 2: Determine Section

Read the existing diary file and decide where to append:

| Content type | Target section | Create if missing | |-------------|----------------|-------------------| | Work experience / technical | ## 2 Work Log | Yes | | Personal note / thought | ## 3 Record | Yes | | User specifies section | As specified | Yes |

If the diary file has no section structure, append at the end.

Step 3: Append Content

Write the content provided by the user (or summarized from the conversation) to the target section.

Universal rules

• **Diary is for the writer's future self, not for explaining things they already

know.** Leave hooks (keywords, dates, context references) instead of unfolding the full story. Aim for "reminder" not "article".

• Filter for human-side content, not agent-side debugging. Diary entries

should capture the human writer's strategic decisions, principles they articulated, surprises in their own thinking, or experience-level reflections. Agent-side tooling pitfalls (broken pre-commit hooks, MCP API gaps, sandbox SSL hiccups, missed git add, plugin install commands) are operational debugging notes — they belong in agent memory (feedback_*.md) or workflow READMEs, NOT in the user's diary. When summarizing a session, ask: "would the human want to recall this in 3 months, or is this just what the agent had to fight through to get the job done?"

• Avoid narrator / reflective framing. Phrases like 「今天最大的洞察」

「之前一直以为……但其实」「借这个机会」「值得记录的是」 are AI-style article openings; strip them. Same in English: "The key insight today is", "What I realized is".

• Prefer prose paragraphs over sub-headings inside an entry. Sub-headings

invite filler (an AI will pad bullets to balance them even when there's nothing to say). One H3 per entry is enough; the body should flow as prose. If multiple logical units exist in one entry, separate them with an inline divider (the user's preferred divider lives in ## 写作风格) rather than another heading.

• Use bullet points or short paragraphs.
• Preserve the user's voice — don't over-polish.
• Add a timestamp prefix if multiple entries exist in the same section:

**14:30** — content here

• If the user provides raw conversation context ("record what we just did"),

summarize the key takeaways concisely.

Style precedence

When the diary content includes Chinese prose, apply this layered precedence (higher layers override lower layers on conflict):

1. tech-doc-style-chinese skill rules (highest) — punctuation (「」

not “”, no exclamation marks), 中英留白, term capitalization (GitHub / Codeup / OpenSpec etc. per its extended list), blacklist words (兜底 / 赋能 / 抓手 …), error-translation table, register guidance (「克制、直接、可执行」). If the tech-doc-style-chinese skill is not loaded in the current session, still apply its core conventions where reasonable.

1. Universal rules above (this skill's own rules: reminder hooks,

no narrator framing, agent-side filter, prose over sub-headings).

1. Personal style overlay from ## 写作风格 in user-config — bold-as-TLDR,

divider character, quote style, em-dash policy, register choice. Apply only for items that don't conflict with layer 1 or 2.

Common conflict resolutions (tech-doc-style-chinese wins):

• register: 工程师口语 (user-config) vs 「克制、直接、可执行」 (tech-doc):

prefer the more restrained tech-doc register; technical terms can still stay in English (e.g., provider, callback, submodule).

• self_deprecating_close: true (user-config): minimize. Tech-doc avoids

decorative closes; if a closing remark is genuinely useful, keep it short and factual rather than self-deprecating (（笑）, （埋雷） etc.).

• parenthetical_asides: true (user-config): allow brief technical

clarifications in parens, but avoid narrator-voice asides.

• quote_style, em_dash, bold_as_tldr, section_divider: typically

aligned across both layers, no conflict.

Step 4: Optional Anki

If the content contains a TIL (Today I Learned) pattern:

"Want to generate Anki cards from this? (invoke anki-card-generator)"

Only ask, don't auto-invoke.

Common Issues

| Issue | Fix | |-------|-----| | Diary path not found | Ask user, suggest saving to ~/.config/nini-skill/diary/user-config.md | | File encoding issues | Obsidian uses UTF-8, ensure write uses UTF-8 | | Section heading mismatch | Read file first, match existing section names |

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: niracler
• Source: niracler/skill
• License: MIT

Install and usage instructions live in the source repository linked above.