Lark Safe Write

A safe-write workflow Skill for Lark (Feishu) wiki. Enforces a four-step loop — preflight, backup, write, verify — so a wiki page never gets clobbered or silently fails to update.

When to use it

• Creating a new wiki node under a Lark knowledge space
• Updating or overwriting an existing wiki page
• Worried that a direct write will destroy existing content
• Need to confirm the page really landed the way you intended

How to trigger it

Say something like:

• "Use lark-safe-write to update this wiki page"
• "Write to the Lark wiki, back it up first and then verify"
• "Create a new Lark wiki doc for me"
• "Overwrite this wiki page, and make a backup first"

What you get back

• Preflight results: bot permissions, token resolution, doc-type check
• Backup path: .backups/YYYYMMDD-HHMMSS-{token}.md
• Write result: success or failure, plus the doc URL
• Verification report: title, first paragraph, and key-marker diff

Problems it solves

When you write into a Lark wiki through lark-cli, these are the traps this Skill closes off:

• Page silently clobbered: A raw update wipes existing content and there is no undo.
• Success that isn't: The API returns 200 but the page never actually changes.
• Path written as content: Passing --markdown /path/to/file.md ends up storing the literal path string instead of the file's contents.
• Permission whack-a-mole: No preflight, so you burn several failed calls before realizing the bot lacks access.

The Skill wraps every write in a mandatory workflow so those failure modes cannot happen in the first place.

Install

Claude Code

# 1. Clone into the local skills directory
git clone https://github.com/duo-lark/lark-safe-write.git ~/.claude/skills/lark-safe-write
 
# 2. Trigger the Skill
/lark-safe-write

OpenAI Codex

# 1. Clone into the Codex skills directory
git clone https://github.com/duo-lark/lark-safe-write.git ~/.agents/skills/lark-safe-write
 
# 2. Restart Codex so the new Skill is picked up
# 3. To trigger: mention the skill name in chat, or reference $lark-safe-write
Use lark-safe-write to update the Lark doc

OpenClaw

# 1. Clone into the OpenClaw skills directory
git clone https://github.com/duo-lark/lark-safe-write.git ~/.openclaw/skills/lark-safe-write
 
# 2. To trigger: reference the skill name in a task or chat
Please use the lark-safe-write flow to write to the wiki

Hermess

# 1. Clone into the Hermess skills directory
git clone https://github.com/duo-lark/lark-safe-write.git ~/.hermes/skills/lark-safe-write
 
# 2. To trigger: just ask in chat
Use the lark-safe-write safe-write flow for this wiki update

Claude Chat / Claude CoWork

Upload SKILL.md straight into the conversation and it is ready to use.

Core workflow

Preflight → Backup → Write → Verify

1. Preflight (cannot be skipped)

• Confirm the --bot identity flag is set
• Run lark-cli auth test to check the token is valid
• Resolve the wiki_token / node_token
• On first touch of a space, confirm the bot actually has access

2. Backup (mandatory on update / overwrite)

• Read the full current content of the doc
• Save it to .backups/YYYYMMDD-HHMMSS-{token}.md
• Report the backup path back to the user

3. Write

• New doc: pin down parent_node_token, then use lark-cli wiki +node-create
• Updating an existing doc: pipe content via stdin (never --markdown /path)
• Wait 2 seconds after the write for Lark's async indexer to catch up

4. Verify (cannot be skipped)

• Re-read the doc
• Diff title, first paragraph, and key markers against what you wrote
• Report the verification result and the doc URL explicitly

Error handling

Dependencies

• lark-cli — bot identity and permissions already configured

License

MIT License