daikenja
SolidNot an assistant — a sage you consult in the shape of a Claude Code plugin. It knows your work, the work around you, and where to go next.
Install
Plugins install via a marketplace, in two steps.
This plugin isn't listed in a marketplace we've indexed. Install it directly from its GitHub repository — the README has the setup steps.
View on GitHubBundles
Everything this plugin ships — skills, agents, commands, hooks, and MCP servers it bundles.
Skills (15)
compose
Rewrites or drafts a work message (Slack, Teams, email) so it stays clear, calm and easy to read for a non-native English audience, without changing the ask, the stance or the confidence level. Returns one balanced message, then offers a firmer or softer version on request. Use this whenever the user pastes a draft message, asks for help replying to a colleague, asks to "make this sound better", "tone this down", "make this sound better before I send it", or has just finished the /daikenja:thread skill and is ready to draft -- even a plain "go ahead" or "draft it" once that skill's intent block exists.
doc-review
Reviews a document against a fixed checklist before it is published or shared -- clarity, undefined terms, unenforced rules, undated decisions, missing owners, contradictions, and content that is stale on its face. Use when the user says "review this doc", "is this page clear", "check this page before I publish", "does this make sense to someone outside the team", or pastes a document/link and asks for a check before sending it out. Read-only -- this skill reports findings, it never edits or rewrites the document. Not for a message or reply (that is /daikenja:compose) or a pre-send substance check on a draft message (that is /daikenja:preflight) -- this skill is for standalone documents (specs, runbooks, wikis, READMEs), not chat replies.
learn-voice
Derives a writing-style.md proposal from writing samples you supply -- your own sent messages, a mail export, a folder of exported threads, a text dump -- and shows the full proposed file before anything is written. Use when the user says "learn how I write", "build my writing style from my old messages", "work out my voice from these", or asks why their writing-style.md is still empty. It reads only samples the user states are their own writing, records style rules and never facts about people or projects, and writes nothing without approval of the exact content. On a file that already holds anything it shows a diff instead of overwriting. Run explicitly with /daikenja:learn-voice -- it never fires on its own. Not for recording how somebody else likes to be written to (that is /daikenja:remember-persona) and not for the rest of first-time setup (that is /daikenja:setup-user).
meeting-review
Turns a meeting transcript into proposed ledger entries -- decisions actually made, action items with owners and dates, and questions left unresolved. Use when the user says "review this meeting", "what came out of this call", "turn this transcript into actions", or pastes a transcript, meeting notes or a recording link and asks what was decided or who owes what. This skill writes nothing itself. It classifies, then hands the entries to the /daikenja:project-log skill, which shows the exact lines and waits for approval. Not for a chat thread (that is /daikenja:thread) and not for coaching on how the user handled the meeting (that is /daikenja:self-review).
preflight
Challenges a draft before it goes out and hands back a revised version plus the facts only you can supply. Runs the six substance checks, then puts the draft in front of a set of reviewer personas -- the busy reader, the executive, the risk reader, a named recipient you describe -- fixes the wording problems they raise, and asks you about anything that needs a fact the draft does not contain. Use for "would this survive X", "poke holes in this", "what will they come back with", "is this ready to go", "should I even raise this", or "am I missing something before I send this". Not for making a message read better when nobody needs to challenge it, which is /daikenja:compose. This skill never sends anything.
project-catchup
Reports what changed in a project's Daikenja ledger since the user last checked, then advances the checkpoint on approval. Use when the user says "catch me up", "what changed since I last looked", "what's new", "what did I miss", or "bring me up to speed" -- personal, delta-shaped asks about a project they already know. Not for a first look at a project (that is /daikenja:project-summary) or a lookup of one specific decision (that is /daikenja:project-decisions). This is the only skill that writes last_checkpoint in daikenja.yaml; it never touches ledger content.
project-decisions
Looks up what was decided about a specific topic in a project's Daikenja ledger, including its supersession history. Use when the user says "was this decided", "what did we decide about X", "what's the decision on X", "show me D-003", or "did we ever settle X" -- a targeted question about one decision or topic, not the whole project. Not for a full project overview (that is /daikenja:project-summary) or a delta since last time (that is /daikenja:project-catchup). Read-only; writes nothing.
project-gaps
Audits a project's Daikenja ledger for open items with no owner or that have sat too long. Use when the user says "what's still open", "what needs an owner", "what's stale", "what's falling through the cracks", "audit the open items", or "what should we be worried about". Not for a full project overview (that is /daikenja:project-summary) or a delta since last time (that is /daikenja:project-catchup). An unowned decision is never reported here -- only Open items are in scope. Read-only; writes nothing.
project-log
Records decisions and open items in a project's Daikenja ledger. Use when the user says "log this", "record this decision", "add this to the ledger", "capture the open items", "note that we agreed X", or pastes a thread or a plain description and asks for what was settled to be written down. Not for a meeting transcript -- that is /daikenja:meeting-review, which classifies it in two passes before handing entries to this skill. Also use when a project has no ledger yet and one is asked for. This is the only skill that writes ledger content -- every other Daikenja skill reads it. A short fact the user dictates is written in the same turn and shown verbatim; everything else is proposed first and written only on approval.
project-summary
Gives a full-state overview of a project's Daikenja ledger, written for someone with no prior context. Use when the user says "what's the state of this project", "give me an overview", "summarize this project", "brief a new person on this", "where do things stand", or is opening a project they have not looked at before. Not for a personal delta since last time (that is /daikenja:project-catchup) or a single decision lookup (that is /daikenja:project-decisions). Read-only; writes nothing.
remember-persona
Records what the user says about a person they write to, in their own personas file (by default ~/.claude/daikenja/personas.md, or a Google Drive file if that is where they keep it), so later messages are written for that reader. Use when the user says "remember that S challenges every technical claim", "log this persona", "note that D is the one who cares about cost", "remember how M likes to be written to", or describes a recipient while drafting and wants that kept. Also the skill other Daikenja skills route through when a description of a person comes up mid-draft. This is the only skill that writes persona content -- every other Daikenja skill reads it, and it scaffolds personas.md from the template on first use if it is not already there. It records only what the user actually said and never infers a character study. Not for a project decision or an open item (that is /daikenja:project-log) and not for the rest of first-time setup, which this skill does not perform (that is /daikenja:setup-user).
self-review
Reviews how the user themselves handled a thread they took part in, and gives private, direct, evidence-backed coaching on their own moves. Use when the user says "how did I handle that", "review my messages in this thread", "what could I have done better", "was I out of line", or pastes a thread they participated in and asks for feedback on their part in it. Findings are about the user alone, never about anyone else's failings, and nothing is written, logged or sent. Not a review of a document (that is /daikenja:doc-review), not a review of what a meeting settled (that is /daikenja:meeting-review), and not a pre-send check on a draft message (that is /daikenja:preflight).
Show all 15 bundled skills Showing all 15 bundled skills
setup-project
Registers the project you are in with Daikenja, sets its per-project settings, and optionally seeds its ledger from sources the project already has. Use when the user says "register this project", "add this repo to Daikenja", "set this project up", or "seed the ledger from our Confluence space". Run explicitly with /daikenja:setup-project -- it never fires on its own. Personal setup stays in /daikenja:setup-user and is not repeated here. Seeding writes nothing itself -- proposed entries go to /daikenja:project-log, which shows the exact lines and waits for approval.
setup-user
One-time (and re-runnable) personal setup for Daikenja. Checks that the session is Claude Code, creates ~/.claude/daikenja/daikenja.yaml from the template, captures the user's profile, copies the blank persona and writing-style files if they are not already there (or creates them in Google Drive if you ask), and reports which connected tools the other skills can use. It is also the only place an upgrade is applied -- when the version recorded in your configuration is behind the installed one, it proposes the edits docs/upgrading.md lists for the versions in between and writes them on approval, which is what the one-line version notice in every other skill points at. It does not register a project -- /daikenja:setup-project does that, and this skill hands off to it at the end, so a first-ever run is still one continuous flow. Run explicitly with /daikenja:setup-user -- it never fires on its own.
thread
Reads a Slack or email thread, summarizes what is being asked and by whom, then collects context from the user before any reply is drafted. Use whenever the user pastes a Slack permalink or thread link, forwards an email thread, pastes a block of chat history, or says things like "what is this about", "help me answer this", "I need to reply to this thread". This skill gathers only -- it never writes the reply itself. The reply is produced later by the /daikenja:compose skill.
Quality Score: 65/100
Details
- Author
- by-carlos
- Repository
- by-carlos/daikenja
- Created
- 6 days ago
- Last Updated
- today
- Language
- Python
- License
- MIT