A pane with the project's open questions (DECISIONS.json), most urgent first, with buttons to answer one directly.

A small collection of reusable SKILL.md instructions for coding agents. The skills are plain Markdown, version-controlled, and portable across compatible agent tools.
This README is the reference: what the skills are, which platforms they support, and how to install them. The [wiki][wiki] answers the other question — why the project is built this way: why a Markdown repository contains Rust crates, why the dev shell compiles bash 3.2 from source, which design choices were made and what they cost, how to verify any of it yourself, and what does not work yet.
[wiki]: https://github.com/tschallacka/ai-skills/wiki
Pick whichever fits how you work. Any of these opens the same interactive installer, where you choose the skills and the agent destination(s).
npx — installs and runs in one step, nothing left on PATH afterward:
npx --yes --package @tschallacka/ai-skills ai-skills-install
Linux / macOS — the one-command installer, no npm required:
curl -fsSL https://raw.githubusercontent.com/tschallacka/ai-skills/master/installer/bootstrap.sh | sh
Windows — the identical command, run inside Git Bash or WSL2 (both provide the POSIX sh it needs; there is no separate PowerShell/cmd installer):
curl -fsSL https://raw.githubusercontent.com/tschallacka/ai-skills/master/installer/bootstrap.sh | sh
See "One-command installer" below for the full list of install destinations, and "npm installation" for a persistent ai-skills-install command on PATH instead of npx.
| Skill | Purpose | Documentation |
|---|---|---|
| Planning | Durable, resumable plans: goals, ordered steps, verification, progress trackers, handoff notes — plus a single-file HTML overview of any plan, and a live served version that updates in place. Steps complete with git-diff atomicity evidence, not decorative checkboxes. | docs |
| Bug report | A defect register in one JSON file: every entry carries its reproduction, observed vs expected, the mechanism, and the verification that fails without the fix. bugs add/bugs update write through shared validation; a closure without proof is refused. | docs |
| Todo | A work queue that outlives the conversation, in one JSON file: nested tasks, every closed item carries its evidence. todo add/todo update keep the register sound; read recipes print user-ready output. | docs |
| Brainstorm | Shapes an under-specified idea into a recorded, agreed picture (brainstorm.md) before planning, with an adversarial completion pass and a plan-vs-implement gate. | docs |
| Post-implementation review | After-the-fact review of built code with concrete proposed fixes, in three passes: implementer self-analysis, an independent solutions agent, and a critical-feedback agent that ranks every fix. | docs |
| Project-specific deviations | Records confirmed project behavior and environment quirks in per-project notes that future agents load instead of re-debugging. | docs |
| Resource-limited testing | Runs heavyweight commands (suites, builds, analyzers, browsers) under platform-appropriate CPU/memory caps, with honest degradation when a platform has no cap mechanism. | docs |
| Chat | RFC-1459 IRC-over-TLS message bus for agents: a rust server a standard TLS IRC client can join, a rust client with UDP discovery and TOFU cert pinning, channels, and additive history/delta reads. | docs |
| Interactive shell | Operates a full-screen terminal program an agent has never seen - nano, mc, lynx, a pager, a menu: a rust PTY wrapper publishing each screen change as one JSONL event, compact row views and deltas to keep context small, element discovery, and a unix-socket client for keys, combos, pastes, mouse and resize. POSIX only. | docs |
| Git worktrees | Parallel agents in one repository: isolated worktree verification, per-agent trees, and merging back in a conflict-aware order without trampling the main checkout. | docs |
| Git merge resolving | Conflicts resolved by what each side changed rather than by ours/theirs: reading intent from history, unions that look like choices, regenerated output, and attributing post-merge failures to the side that caused them. | docs |
| Merge request etiquette | Descriptions a reviewer can act on, in the author's voice: a one-paragraph TLDR, the defect/cause/change body, derived from the branch's commits, and the one case where a collapsible section earns its place. | docs |
| Text etiquette | Shorthand and a clipped register for an agent's prose - chat, dev talk, and its own thinking: facts first, a shared shorthand with an ask-don't-guess rule, praise capped at gj, and the people-please prose banned. Plain english on request. | docs |
| Question etiquette | Numbered questions, lettered options, never a bullet: a reply like Q7b is unambiguous, a partial answer names exactly which numbers are still open, and a lettered list always ends with "none of these, I'll say it myself." | docs |
| AI text editor | Server-owned agent editor tabs with explicit search, revision-aware edits, undo/redo, raw-byte and hex modes, SQLite metadata, and Unix/TCP transport. | docs |
| www | A brake the human can pull, and one the agent pulls on itself when it is thrashing: stop, answer what do we have / what are the values / what are we trying to achieve, in order, then continue with one reasoned step or a numbered question. | docs |
| CI failures | What actually failed in a CI run or pipeline, from a run/pipeline id, a PR/MR number, or a branch: GitHub and GitLab detected from the git remote, named rather than chosen silently, with just the failing lines extracted per job. | docs |
| Decisions | A register of non-blocking questions raised mid-work: Q# ids, lettered options, priority and the branch they came from as context, so a question can be stubbed and left open without blocking a turn. | docs |
| rjq | Parses, filters and searches JSON with the shipped rjq binary, a jq-compatible tool, on machines without jq. | docs |
| Tailpipe | A less/tail for agents: pipe a command's output into a named, server-held stream with chat-style message ids; a reader lists/reads/searches/tails it from anywhere, an MCP adapter offers the same as typed tools, and a board mod shows it to a human. Idle streams are gzip-snapshotted after 15 minutes. | docs |
Use a skill only when its frontmatter trigger matches the task or when the user explicitly requests it. Each skill documents when not to activate.
"Portable" elsewhere in this repository means portable across agent tools (Claude Code, Codex, OpenCode, OpenClaw, Cline). Operating-system support is separate and stated here:
Those tools do not identify the calling agent the same way, which matters to every per-agent feature here. src/agent-session-key/HARNESS-IDENTITY.md records what each one provides, how it was measured, and the procedure for contributing a harness that is not yet listed.
| Supported | |
|---|---|
| Linux | any distribution, bash 4 or 5, GNU userland |
| macOS | 11+ with the stock /bin/bash 3.2, BSD userland; Homebrew bash not required |
| Windows | Git for Windows' bash with its bundled coreutils, checked by CI legs, or WSL2, which is a Linux install. The evidence and the conventions are in .agents/MAINTAINER.md 1.16, which is in a full git checkout and not part of the installed package |
The one-command install needs a POSIX sh (bootstrap.sh has no bash-only constructs), plus curl, tar, awk, and standard coreutils; the installed skills' own helper scripts need bash, POSIX coreutils, awk, sed, grep, and git. The planning skill additionally needs rjq, and on macOS resource-limited-testing needs memlimit; the installer checks for both up front and prints a per-platform install hint rather than failing partway through. Those two are the only extra runtime dependencies any skill has — in particular python3 is not required by anything that gets installed, only by this repository's own benchmark harness. CODE-STYLE.md is the contract these scripts are held to; what CI proves on each platform is mapped in .agents/MAINTAINER.md section 3 (a full git checkout has both; the installed package has neither).
One skill is genuinely OS-scoped: resource-limited-testing enforces a hard RAM cap only on Linux, via a transient systemd --user cgroup v2 scope. On macOS it uses memlimit (MIT, by Jelle Besseling), which refuses allocations past the cap instead of killing the process — a best-effort cap on real resident memory over the process tree, not a cgroup-equivalent guarantee; SKILL.md lists what it does and does not promise. On Apple Silicon macOS the installer declares memlimit a soft requirement: without it the skill still installs and the run warns that the RAM cap is not enforced. The wrapper then degrades to nice plus cpulimit (CPU throttling only); on a Linux session without a user systemd instance it falls back to ulimit -v. memlimit is not asked for on Intel Macs at all — it does not support them. Every other skill behaves identically on both.
Run this command and choose the skills and agent destination interactively:
curl -fsSL https://raw.githubusercontent.com/tschallacka/ai-skills/master/installer/bootstrap.sh | sh
The installer can install all the skills or one skill, and supports these global skill roots:
| Destination | Agent or standard |
|---|---|
~/.agents/skills | Universal Agent Skills root; recommended shared destination |
~/.codex/skills | Codex CLI |
~/.claude/skills | Claude Code |
~/.config/opencode/skills | OpenCode |
~/.openclaw/skills | OpenClaw managed skills |
~/.cline/skills | Cline |
The universal root is also discovered by OpenCode and OpenClaw. Installing the same skill into multiple roots can create duplicate definitions or precedence conflicts, so choose only the roots you need.
The interactive installer checks which supported agents are present and omits roots for agents it cannot detect. Custom roots are saved in ~/.config/tsch-ai-skills/custom-locations and are offered again when they still exist.
Install the package globally to expose the installer command. The npm package keeps the skills in this repository and links ai-skills-install directly to installer/bootstrap.sh, which fetches the matching compiled installer release for your platform on first run:
npm install -g @tschallacka/ai-skills
ai-skills-install
For a one-off run without a global install:
npx --yes --package @tschallacka/ai-skills ai-skills-install
The npm package does not install skills automatically as an npm lifecycle side-effect; run the installer command when you are ready to choose a target.
Run the installer again against the same root. It compares every managed file with the repository copy and reports one of three outcomes per destination: Up to date (nothing differed), Installed (files were written), or Skipped (you declined, or the destination needs manual review).
Each installed skill contains a .version marker identifying its tag, branch, and commit. If an installed file differs from the repository version, the installer asks before replacing it. For a managed version transition that the user approves — the .version marker differs, so the change came from a new release rather than from you — old files are replaced without backups; the previous version can be restored by running the installer against its tag with AI_SKILLS_REF. Unmanaged changes still receive <file>.bak backups (.bak.1, .bak.2, … if one already exists). A symlinked skill is skipped for manual review rather than following the link and modifying an unexpected location.
Interactively, choose that one skill at the menu. Headless, name it:
curl -fsSL https://raw.githubusercontent.com/tschallacka/ai-skills/master/installer/bootstrap.sh \
| sh -s -- install --skill planning --target "$HOME/.codex/skills"
--skill may be given more than once, and each value may itself be a comma-separated list, so --skill planning --skill brainstorm and --skill planning,brainstorm install the same two. Repeats are collapsed, menu numbers work (--skill 1,4), and all selects everything wherever it appears.
Use --all, --skill, --target and --yes when the choices are already known. --target takes a single root, so installing into two roots is two runs.
# Install all skills into the shared Agent Skills root
curl -fsSL https://raw.githubusercontent.com/tschallacka/ai-skills/master/installer/bootstrap.sh \
| sh -s -- install --all --target "$HOME/.agents/skills"
# Unattended replacement: managed version transitions replace without backups,
# unmanaged changed files are still backed up as <file>.bak
curl -fsSL https://raw.githubusercontent.com/tschallacka/ai-skills/master/installer/bootstrap.sh \
| sh -s -- install --all --target "$HOME/.agents/skills" --yes
Every run ends with a summary block on stdout saying what was installed, what was not, and why; the progress and diagnostics go to stderr, so installer install … > summary.txt keeps the outcome and 2>/dev/null keeps it readable. A blocked skill is reported once — not once per root — with the commands that finish the job. This is what an --all run on a machine without rjq prints:
== Summary ==
Installed: /home/u/.agents/skills/project-specifics
Installed: /home/u/.agents/skills/resource-limited-testing
Installed: /home/u/.agents/skills/brainstorm
Installed: /home/u/.agents/skills/post-implementation-review
Skipped: planning — a hard requirement is missing, nothing was written
To install planning once its requirements are met:
1. install rjq:
sudo apt-get install -y rjq
2. replay this run:
installer install --skill planning --target /home/u/.agents/skills --yes
The install step is chosen for the detected platform and package manager, and the replay line carries the same target and flags as the run that printed it, naming the installer binary bootstrap.sh downloaded to run it. The exit status is non-zero, because four of five skills is a partial install and CI must not read it as success.
Dependencies are declared per skill, so one unsatisfiable dependency never stops the other skills from installing. Each skill ships a requires.tsv naming what it needs, on which platform and architecture, and how badly:
| Strength | Effect |
|---|---|
hard | The skill does not work without the tool. It is not installed, the run explains why and prints the replay commands, and the exit status is non-zero. |
soft | The skill works in a degraded form. It is installed, with a warning naming the tool and the capability that is lost. The exit status is unaffected. |
Currently:
planning requires rjq (hard, every platform) — without it validate-plan.sh refuses to run and the plan gates stop firing.resource-limited-testing names memlimit (soft, Apple Silicon macOS only) — see Supported platforms above for what the degraded path still does.No other skill has a runtime dependency.
| Code | Meaning |
|---|---|
| 0 | Everything requested was installed. Soft warnings do not change this. |
| 1 | A requested skill was blocked by a hard requirement, or any other error. |
| 2 | install-skill only: approval declined, nothing was written. |
| 3 | install-skill only: an unsafe collision (an existing file that is not a managed version upgrade, or a symlink). |
Codes 2 and 3 belong to the machine-facing install-skill subcommand that the planning skill's own tooling uses; the interactive, install --all, and install --skill paths only ever return 0 or 1.
Running the bare one-liner with no arguments, or installer interactive directly, opens a full-screen skill picker instead of the numbered menu:
| Key | Action |
|---|---|
↑/k, ↓/j, PageUp, PageDown, Home, End | move the cursor |
Enter / Space | toggle the skill under the cursor |
Tab / Shift-Tab | switch focus between the skill list and the info pane |
a / n | select all / select none |
d, r, m (info pane focused) | show dependency hints, re-verify requirements, cycle a skill's integration mode |
i | confirm and install the current selection |
q / Escape | quit without installing |
With neither --target nor --agent given, it also prompts to choose an auto-detected agent root, a saved custom directory, a new custom directory, or a for every listed root.
Review the installer before running it if you do not trust the source. Skills are instructions that may guide agents to run commands or access files.
install.sh retired in favor of a compiled Rust installer (src/installer/); installer/bootstrap.sh is the small, pure-POSIX-sh entry point (#!/usr/bin/env sh, no bash-only constructs) that detects the platform, downloads the matching release, and hands off to it — see CONTRIBUTING.md before editing either.
installer/bootstrap.sh always downloads a release archive, so it is not how a checkout installs its own local files. Build the installer and point it at the checkout with --source instead:
cargo build --release -p installer
./target/release/installer interactive --source .
--package dev (also read from $PACKAGE_SELECTION) ships the MODE: DEV files too — tests, maintainer docs — instead of filtering them out, for installing a working development copy rather than the prod set.
installer/bootstrap.sh itself accepts AI_SKILLS_REPO_URL (a different owner/repo to resolve GitHub's "latest release" redirect against) and AI_SKILLS_RELEASE_URL (an exact archive URL, bypassing that redirect entirely — how RELEASE.md verifies one specific tag).
benchmark/planning/runtime/ makes the worker/reviewer/analyzer launch, session-id extraction, and token telemetry pluggable per CLI: the active agent defaults to codex and is selected with BENCHMARK_AGENT (opencode, claude), with a shared launcher (lib-agent.sh) owning all setsid/timeout/process-group control. See benchmark/planning/runtime/README.md for the contract and first-time setup.resource-limited-testing's platform behaviour is described under Supported platforms; its SKILL.md documents each fallback in detail.Distributed under the MIT License.
hooks/register.tsx 543 lines1import type { Register } from 'claude-code'
2import { update } from 'claude-code'
3
4// The project's questions (DECISIONS.json), most urgent first. A question's
5// lifecycle is open -> decided -> implemented: the user answering one does
6// not make it vanish from this pane, since a decided question is still
7// outstanding work -- the agent's, not the user's -- until it is actually
8// carried out and marked implemented. The pane's View toggle has one button
9// per status -- Open (needs the user's pick), Pending (decided, awaiting
10// the agent's implementation), Implemented (already done) -- and defaults
11// to Open. The register is read on every draw, the same way
12// register-board reads BUGS.json/TODO.json, so the pane always shows what is
13// on disk; a timer also forces a redraw, the same pattern ci-board's
14// JOB_REFRESH_MS uses, so a question answered or implemented elsewhere (the
15// CLI, or an agent via MCP) updates without the person reopening the pane.
16// The interval is short (REFRESH_MS below): unlike ci-board's network-bound
17// GitHub calls, this is a single cheap local file read, so there is no
18// reason to make a person wait up to 30 seconds to see a question raised a
19// moment ago.
20// Toggled by `enabled` in settings.json pluginConfigs["decision-board"].options.
21
22const PANE = 'decision-board'
23const TOOL = 'show_decision_board'
24const READ = 'read_decision_board'
25const ANSWER = 'answer_decision_board'
26const IMPLEMENT = 'implement_decision_board'
27
28const priorityFilterKey = { plugin: 'decision-board', key: 'priorityFilter' } as const
29const branchOnlyKey = { plugin: 'decision-board', key: 'branchOnly' } as const
30const viewKey = { plugin: 'decision-board', key: 'view' } as const
31const lastLoggedKey = { plugin: 'decision-board', key: 'lastLogged' } as const
32
33// How long the "decision logged" banner stays up after a person answers a
34// question from the pane, in milliseconds -- long enough to read and act on,
35// short enough that it is gone well before anyone forgets why it was there.
36// It piggybacks on the pane's own REFRESH_MS redraw rather than a timer of
37// its own, so it fades on the next tick past this window.
38const BANNER_MS = 8_000
39
40// The part of `$.fs` the reading needs.
41type Fs = {
42 exists: (path: string) => Promise<boolean>
43 read: (path: string) => Promise<string>
44}
45
46// The part of `$.process.run` the resolution needs.
47type ProcessRun = (argv: readonly string[]) => Promise<{ exitCode: number; stdout: string; stderr: string }>
48
49type Choice = { letter: string; label: string }
50type Question = {
51 id: string
52 title: string
53 status: string
54 priority: string
55 branch: string
56 options: Choice[]
57 context: string
58 chosen: string | null
59 resolution: string | null
60 created_at: string
61 updated_at: string
62}
63
64type View = 'open' | 'pending' | 'implemented'
65
66const PRIORITY_ORDER = ['urgent', 'high', 'normal', 'low', 'someday']
67
68// One color per priority, so the list reads at a glance without counting on
69// bold alone; a priority this register does not know about (future-proofing
70// against a schema addition) falls back to no color rather than guessing.
71const PRIORITY_COLOR: Record<string, string> = {
72 urgent: 'red',
73 high: 'yellow',
74 normal: 'cyan',
75 low: 'blue',
76 someday: 'gray',
77}
78
79function rank(value: string): number {
80 const index = PRIORITY_ORDER.indexOf(value)
81 return index === -1 ? PRIORITY_ORDER.length : index
82}
83
84// Urgent first; `Array.prototype.sort` is a stable sort, so two questions of
85// the same priority keep the order the register itself lists them in.
86function sortQuestionsUrgentFirst(questions: Question[]): Question[] {
87 return [...questions].sort((a, b) => rank(a.priority) - rank(b.priority))
88}
89
90// Open and decided are both still outstanding (an answer is not the same as
91// an implementation); implemented, closed, dropped and obsolete are resting
92// states this pane does not otherwise surface. `pendingQuestions` is the
93// agent-facing tools' own grouping (what still needs someone's attention,
94// open or decided alike); the pane's View toggle instead shows exactly one
95// status at a time, via `questionsForView`.
96function pendingQuestions(questions: Question[]): Question[] {
97 return questions.filter(q => q.status === 'open' || q.status === 'decided')
98}
99
100function openQuestions(questions: Question[]): Question[] {
101 return questions.filter(q => q.status === 'open')
102}
103
104function decidedQuestions(questions: Question[]): Question[] {
105 return questions.filter(q => q.status === 'decided')
106}
107
108function implementedQuestions(questions: Question[]): Question[] {
109 return questions.filter(q => q.status === 'implemented')
110}
111
112function questionsForView(questions: Question[], view: View): Question[] {
113 if (view === 'implemented') return implementedQuestions(questions)
114 if (view === 'pending') return decidedQuestions(questions)
115 return openQuestions(questions)
116}
117
118// The pane's own priority/branch filters, each optional and independent of
119// the other. Narrowing is the board's own convenience: the agent-facing
120// tools (show/read) never apply it, so an agent always sees the whole truth
121// regardless of what a person last picked in the pane.
122function filterQuestions(
123 questions: Question[],
124 filter: { priority: string | null; branch: string | null },
125): Question[] {
126 return questions.filter(
127 q =>
128 (filter.priority === null || q.priority === filter.priority) &&
129 (filter.branch === null || q.branch === filter.branch),
130 )
131}
132
133// Every question in the register, read from its own already-resolved
134// DECISIONS.json (see resolveDecisionsFile), sorted urgent-first. No
135// register file at all reads as no questions, the same as register-board
136// treats a missing BUGS.json/TODO.json.
137async function loadQuestions(fs: Fs, file: string): Promise<Question[]> {
138 if (!(await fs.exists(file))) return []
139 const parsed = JSON.parse(await fs.read(file)) as { questions?: Question[] }
140 return sortQuestionsUrgentFirst(parsed.questions ?? [])
141}
142
143// The decisions binary's path, in the shared bin the installer puts it in --
144// the same resolution ci-board's own toolPath uses for ci-failures.
145function decisionsBin(home: string, xdg: string): string {
146 const bin = xdg ? `${xdg}/tsch-ai-skills/bin` : `${home}/.config/tsch-ai-skills/bin`
147 return `${bin}/decisions`
148}
149
150// The DECISIONS.json path `decisions` itself actually reads and writes,
151// resolved once via `decisions resolve-path` per handler invocation (its
152// result is threaded through to every reader/writer that same invocation
153// needs, rather than re-resolved) -- an absolute worktree path is used as
154// is; the bare "DECISIONS.json" default is joined onto the session root,
155// exactly where this mod read and wrote it before this resolution existed.
156// A process-run failure (the binary missing, for example) falls back the
157// same way.
158async function resolveDecisionsFile(run: ProcessRun, home: string, xdg: string, root: string): Promise<string> {
159 const result = await run([decisionsBin(home, xdg), 'resolve-path']).catch(() => undefined)
160 if (!result || result.exitCode !== 0) return `${root}/DECISIONS.json`
161 const file = result.stdout.trim()
162 if (file === '') return `${root}/DECISIONS.json`
163 return file.startsWith('/') ? file : `${root}/${file}`
164}
165
166// The argument list `decisions answer <id> <letter> --file PATH` expects,
167// built once here so the button handler and the agent-facing tool call the
168// binary the identical way.
169function answerArgs(bin: string, id: string, letter: string, file: string): string[] {
170 return [bin, 'answer', id, letter, '--file', file]
171}
172
173// `decisions implement <id> [note] --file PATH`; note is optional, unlike
174// answer's letter, so an empty one is simply omitted rather than passed as
175// an empty positional.
176function implementArgs(bin: string, id: string, note: string, file: string): string[] {
177 const args = [bin, 'implement', id]
178 if (note) args.push(note)
179 args.push('--file', file)
180 return args
181}
182
183// "Date when asked", in the loose relative phrasing a person reads faster
184// than a timestamp: same-day counts in minutes/hours, this month in days,
185// the next in "last month", older in whole months, then whole years. An
186// unparseable timestamp reads as '' rather than 'NaN years ago'.
187function relativeAge(iso: string, nowMs: number): string {
188 const then = Date.parse(iso)
189 if (!Number.isFinite(then)) return ''
190 const seconds = Math.max(0, Math.round((nowMs - then) / 1000))
191 const minute = 60
192 const hour = 3600
193 const day = 86400
194 const month = 2_592_000 // 30 days
195 const year = 31_536_000 // 365 days
196 if (seconds < minute) return 'just now'
197 if (seconds < hour) {
198 const n = Math.floor(seconds / minute)
199 return `${n} minute${n === 1 ? '' : 's'} ago`
200 }
201 if (seconds < day) {
202 const n = Math.floor(seconds / hour)
203 return `${n} hour${n === 1 ? '' : 's'} ago`
204 }
205 if (seconds < month) {
206 const n = Math.floor(seconds / day)
207 return n === 1 ? 'yesterday' : `${n} days ago`
208 }
209 if (seconds < month * 2) return 'last month'
210 if (seconds < year) return `${Math.floor(seconds / month)} months ago`
211 const n = Math.floor(seconds / year)
212 return n === 1 ? 'last year' : `${n} years ago`
213}
214
215// The label text for whichever option letter was chosen, or the letter
216// itself if the register no longer lists it (an option set edited after the
217// pick, in principle -- defensive rather than expected).
218function chosenLabel(question: Question): string {
219 const picked = question.options.find(o => o.letter === question.chosen)
220 return picked ? `${question.chosen}: ${picked.label}` : question.chosen ?? ''
221}
222
223// Exposes the pure functions above for a unit test to call directly, without
224// spawning the binary or driving the mod runtime -- nothing here is read by
225// Claude Code itself, which only ever reads the `register` export below.
226export const __test = {
227 sortQuestionsUrgentFirst,
228 answerArgs,
229 implementArgs,
230 filterQuestions,
231 pendingQuestions,
232 openQuestions,
233 decidedQuestions,
234 implementedQuestions,
235 questionsForView,
236 relativeAge,
237 chosenLabel,
238 loadQuestions,
239 decisionsBin,
240 resolveDecisionsFile,
241}
242
243export const register: Register = (on, options) => {
244 if (options.enabled === false) return
245
246 on('session.start', async ($, e, next) => {
247 await $.command.register({
248 name: 'decision-board',
249 description: 'Show the project’s questions in a pane; `/decision-board close` hides it',
250 })
251 // questions-board is an alias: the same board, under the name that matches
252 // bugs-board/todo-board's own naming scheme.
253 await $.command.register({
254 name: 'questions-board',
255 description: 'Alias for /decision-board: show the project’s questions in a pane; `/questions-board close` hides it',
256 })
257 await $.tool.register({
258 name: TOOL,
259 description: 'Show the project’s pending (open + decided) questions to the person, in a pane, and return them as text for the chat.',
260 inputSchema: { type: 'object', properties: {} },
261 })
262 await $.tool.register({
263 name: READ,
264 description: 'Read the project’s pending (open + decided) questions as text, without opening the pane.',
265 inputSchema: { type: 'object', properties: {} },
266 })
267 await $.tool.register({
268 name: ANSWER,
269 description: 'Record the person’s pick for an open question: sets it to decided -- pending implementation, not yet done.',
270 inputSchema: {
271 type: 'object',
272 properties: {
273 id: { type: 'string', description: 'The question id, e.g. Q3.' },
274 letter: { type: 'string', description: 'One of the question’s own option letters.' },
275 },
276 required: ['id', 'letter'],
277 },
278 })
279 await $.tool.register({
280 name: IMPLEMENT,
281 description: 'Mark a decided question as carried out in the code, recording what was done. Refused unless the question is currently decided.',
282 inputSchema: {
283 type: 'object',
284 properties: {
285 id: { type: 'string', description: 'The question id, e.g. Q3.' },
286 note: { type: 'string', description: 'What was implemented, for the record.' },
287 },
288 required: ['id'],
289 },
290 })
291 return next(e)
292 })
293
294 on('command.run', { command: 'decision-board' }, async ($, e) => {
295 if (e.args.trim().toLowerCase() === 'close') {
296 await $.ui.close({ id: PANE })
297 return { text: 'Decision board closed.' }
298 }
299 await $.ui.open({ id: PANE, title: 'Questions' })
300 return { text: 'Decision board opened.' }
301 })
302
303 // questions-board is an alias for decision-board, opening the same pane.
304 on('command.run', { command: 'questions-board' }, async ($, e) => {
305 if (e.args.trim().toLowerCase() === 'close') {
306 await $.ui.close({ id: PANE })
307 return { text: 'Decision board closed.' }
308 }
309 await $.ui.open({ id: PANE, title: 'Questions' })
310 return { text: 'Decision board opened.' }
311 })
312
313 // The agent's tools (show/read/answer/implement) deliberately never apply
314 // the pane's own priority/branch/view filters: an agent must always see
315 // the whole truth, regardless of what a person last narrowed the view to.
316 on('tool.call', { tool: `mcp__decision-board__${TOOL}` }, async $ => {
317 const root = await $.session.root()
318 const home = (await $.env.get('HOME')) ?? ''
319 const xdg = (await $.env.get('XDG_CONFIG_HOME')) ?? ''
320 const fs: Fs = { exists: path => $.fs.exists(path), read: path => $.fs.read(path) }
321 const file = await resolveDecisionsFile(argv => $.process.run(argv), home, xdg, root)
322 const all = await loadQuestions(fs, file)
323 const pending = pendingQuestions(all)
324 await $.ui.open({ id: PANE, title: 'Questions' })
325
326 return {
327 result: [
328 `Pending questions: ${pending.length}`,
329 ...pending.map(q => `${q.id} [${q.priority}/${q.status}] ${q.title} (${q.branch})`),
330 `Implemented: ${implementedQuestions(all).length}`,
331 ].join('\n'),
332 }
333 })
334
335 // The agent's read of the pending questions. Read-only; the pane is not opened.
336 on('tool.call', { tool: `mcp__decision-board__${READ}` }, async $ => {
337 const root = await $.session.root()
338 const home = (await $.env.get('HOME')) ?? ''
339 const xdg = (await $.env.get('XDG_CONFIG_HOME')) ?? ''
340 const fs: Fs = { exists: path => $.fs.exists(path), read: path => $.fs.read(path) }
341 const file = await resolveDecisionsFile(argv => $.process.run(argv), home, xdg, root)
342 const all = await loadQuestions(fs, file)
343 const pending = pendingQuestions(all)
344 const live = await $.ui.selection().catch(() => undefined)
345
346 return {
347 result: [
348 `Pending questions: ${pending.length}`,
349 ...pending.map(q => `${q.id} [${q.priority}/${q.status}] ${q.title} (${q.branch})`),
350 `Implemented: ${implementedQuestions(all).length}`,
351 `Highlighted in the board: ${live?.text ?? 'none'}`,
352 ].join('\n'),
353 }
354 })
355
356 // The agent's way to answer a question directly, without opening the pane
357 // or pressing a button -- the same underlying call the pane's own buttons
358 // make (answerArgs), so the two paths can never disagree about what the
359 // binary is told.
360 on('tool.call', { tool: `mcp__decision-board__${ANSWER}` }, async ($, e) => {
361 const root = await $.session.root()
362 const home = (await $.env.get('HOME')) ?? ''
363 const xdg = (await $.env.get('XDG_CONFIG_HOME')) ?? ''
364 const id = String(e.id ?? '').trim()
365 const letter = String(e.letter ?? '').trim()
366 if (!id || !letter) return { result: 'Both id and letter are required.' }
367 const file = await resolveDecisionsFile(argv => $.process.run(argv), home, xdg, root)
368 const bin = decisionsBin(home, xdg)
369 const run = await $.process.run(answerArgs(bin, id, letter, file))
370 $.ui.invalidate('ui.render')
371 return { result: run.exitCode === 0 ? `${id} decided: ${letter}.` : (run.stderr.trim() || `decisions exited ${run.exitCode}`) }
372 })
373
374 // The agent's way to record that a decided question's pick was carried
375 // out. Deliberately agent-only: the pane itself offers no button for this
376 // -- a human picking an option is "decided", not "done", and only the
377 // agent that actually did the work can say when that is true.
378 on('tool.call', { tool: `mcp__decision-board__${IMPLEMENT}` }, async ($, e) => {
379 const root = await $.session.root()
380 const home = (await $.env.get('HOME')) ?? ''
381 const xdg = (await $.env.get('XDG_CONFIG_HOME')) ?? ''
382 const id = String(e.id ?? '').trim()
383 const note = String(e.note ?? '').trim()
384 if (!id) return { result: 'id is required.' }
385 const file = await resolveDecisionsFile(argv => $.process.run(argv), home, xdg, root)
386 const bin = decisionsBin(home, xdg)
387 const run = await $.process.run(implementArgs(bin, id, note, file))
388 $.ui.invalidate('ui.render')
389 return { result: run.exitCode === 0 ? `${id} implemented.` : (run.stderr.trim() || `decisions exited ${run.exitCode}`) }
390 })
391
392 const REFRESH_MS = 2_000
393 let refresh: { cancel: () => void } | undefined
394
395 on('ui.render', { component: 'Pane', requestId: PANE }, async ($, e) => {
396 const { Box, Button, Text } = $.ui.resolve(e)
397 const root = await $.session.root()
398 const home = (await $.env.get('HOME')) ?? ''
399 const xdg = (await $.env.get('XDG_CONFIG_HOME')) ?? ''
400 const fs: Fs = { exists: path => $.fs.exists(path), read: path => $.fs.read(path) }
401 const file = await resolveDecisionsFile(argv => $.process.run(argv), home, xdg, root)
402 const all = await loadQuestions(fs, file)
403
404 if (!refresh) refresh = $.clock.every(REFRESH_MS, () => $.ui.invalidate('ui.render'))
405
406 const { value: priorityFilter } = await $.state.get(priorityFilterKey)
407 const { value: branchOnly } = await $.state.get(branchOnlyKey)
408 const { value: viewValue } = await $.state.get(viewKey)
409 const view: View = viewValue === 'pending' || viewValue === 'implemented' ? viewValue : 'open'
410 const { value: lastLogged } = await $.state.get(lastLoggedKey)
411 const showBanner = !!lastLogged && Date.now() - lastLogged.at < BANNER_MS
412 // Only asks git for the current branch when the toggle actually needs it:
413 // every other render costs one file read and nothing else.
414 let activeBranch: string | null = null
415 if (branchOnly) {
416 const head = await $.process.run(['git', 'rev-parse', '--abbrev-ref', 'HEAD']).catch(() => undefined)
417 activeBranch = head && head.exitCode === 0 ? head.stdout.trim() : null
418 }
419 const inView = questionsForView(all, view)
420 const questions = filterQuestions(inView, { priority: priorityFilter ?? null, branch: activeBranch })
421 const now = Date.now()
422
423 const answer = async (id: string, letter: string) => {
424 // Reuses the `file` resolved once above for this render -- a second
425 // resolve-path call here would double the process cost per press.
426 const bin = decisionsBin(home, xdg)
427 const run = await $.process.run(answerArgs(bin, id, letter, file))
428 if (run.exitCode === 0) {
429 await update($, lastLoggedKey, () => ({ id, at: Date.now() }))
430 }
431 $.ui.invalidate('ui.render')
432 }
433
434 const HEADINGS: Record<View, string> = { open: 'Open questions', pending: 'Pending questions', implemented: 'Implemented questions' }
435 const heading = HEADINGS[view]
436 // Said once per view, not once per row -- a row's own line stays just
437 // "picked X", since repeating the explanation on every question in a
438 // long pending list is noise once the person already knows what the
439 // view means.
440 const SUBTITLES: Partial<Record<View, string>> = {
441 pending: 'Decided; the agent implements these and marks them, not the person.',
442 }
443
444 return (
445 <Box flexDirection="column">
446 <Box flexDirection="row" justifyContent="space-between">
447 <Text bold inverse>
448 {heading}
449 </Text>
450 <Button role="dismiss" onPress={() => $.ui.close({ id: PANE })}>
451 Close
452 </Button>
453 </Box>
454 {SUBTITLES[view] && <Text dimColor>{SUBTITLES[view]}</Text>}
455 {showBanner && (
456 <Text color="#ffa500" bold>
457 {`Decision logged (${lastLogged!.id}). Please remind Claude in chat -- sadly this can't be automated.`}
458 </Text>
459 )}
460 <Text> </Text>
461 <Box flexDirection="row">
462 <Text dimColor>View: </Text>
463 <Button
464 variant={view === 'open' ? 'primary' : undefined}
465 onPress={() => update($, viewKey, () => 'open')}
466 >
467 Open
468 </Button>
469 <Text> </Text>
470 <Button
471 variant={view === 'pending' ? 'primary' : undefined}
472 onPress={() => update($, viewKey, () => 'pending')}
473 >
474 Pending
475 </Button>
476 <Text> </Text>
477 <Button
478 variant={view === 'implemented' ? 'primary' : undefined}
479 onPress={() => update($, viewKey, () => 'implemented')}
480 >
481 Implemented
482 </Button>
483 </Box>
484 <Box flexDirection="row">
485 <Text dimColor>Priority: </Text>
486 {PRIORITY_ORDER.map(p => (
487 <Box key={p} flexDirection="row">
488 <Button
489 variant={priorityFilter === p ? 'primary' : undefined}
490 onPress={() => update($, priorityFilterKey, () => (priorityFilter === p ? null : p))}
491 >
492 {p}
493 </Button>
494 <Text> </Text>
495 </Box>
496 ))}
497 <Button
498 variant={branchOnly ? 'primary' : undefined}
499 onPress={() => update($, branchOnlyKey, () => !branchOnly)}
500 >
501 this branch only
502 </Button>
503 </Box>
504 <Text> </Text>
505 {inView.length === 0 && (
506 <Text dimColor>
507 {{ open: 'No open questions.', pending: 'Nothing decided yet.', implemented: 'Nothing implemented yet.' }[view]}
508 </Text>
509 )}
510 {inView.length > 0 && questions.length === 0 && <Text dimColor>No questions match this filter.</Text>}
511 {questions.map(q => (
512 <Box key={q.id} flexDirection="column">
513 <Text bold color={PRIORITY_COLOR[q.priority]}>
514 {`${q.id} [${q.priority}${q.status === 'decided' ? ' · decided' : ''}] ${q.title}`}
515 </Text>
516 <Text italic dimColor>
517 {` on ${q.branch || 'no branch recorded'} · asked ${relativeAge(q.created_at, now) || 'at an unknown time'}`}
518 </Text>
519 {q.status === 'open' && (
520 <Box flexDirection="column">
521 {q.options.map(option => (
522 <Box key={option.letter} flexDirection="row">
523 <Button onPress={() => answer(q.id, option.letter)}>{`${option.letter}: ${option.label}`}</Button>
524 </Box>
525 ))}
526 </Box>
527 )}
528 {q.status === 'decided' && (
529 <Text dimColor>{` picked ${chosenLabel(q)}`}</Text>
530 )}
531 {q.status === 'implemented' && (
532 <Text dimColor>
533 {` picked ${chosenLabel(q)}${q.resolution ? ` · ${q.resolution}` : ''} · implemented ${relativeAge(q.updated_at, now) || 'at an unknown time'}`}
534 </Text>
535 )}
536 <Text> </Text>
537 </Box>
538 ))}
539 </Box>
540 )
541 })
542}
543types/index.d.ts 22 lines1// The person's own priority/branch filters and Open/Pending/Implemented view
2// toggle for the pane, kept across redraws the same way ci-board keeps its
3// own browsing state: null means "unset, default to open" for `view`,
4// the same way null/false means "no filter" for the others -- the pane
5// always has a definite state to draw from. `register-board`'s own
6// types/index.d.ts has nothing to augment here (it keeps no `$.state` at
7// all); this mod now does.
8declare module 'claude-code' {
9 interface PluginState {
10 'decision-board': {
11 priorityFilter: string | null
12 branchOnly: boolean
13 view: 'open' | 'pending' | 'implemented' | null
14 lastLogged: { id: string; at: number } | null
15 }
16 }
17 interface McpToolInputs {
18 'mcp__decision-board__answer_decision_board': { id: string; letter: string }
19 'mcp__decision-board__implement_decision_board': { id: string; note?: string }
20 }
21}
22