@installagent/sdk 0.1.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/dist/agent-templates.d.ts +90 -0
- package/dist/agent-templates.js +211 -0
- package/dist/client-build.d.ts +50 -0
- package/dist/client-build.js +155 -0
- package/dist/client-context.d.ts +40 -0
- package/dist/client-context.js +70 -0
- package/dist/client-drive.d.ts +67 -0
- package/dist/client-drive.js +278 -0
- package/dist/client-intake.d.ts +68 -0
- package/dist/client-intake.js +189 -0
- package/dist/client-linear.d.ts +30 -0
- package/dist/client-linear.js +85 -0
- package/dist/docs-search.d.ts +44 -0
- package/dist/docs-search.js +187 -0
- package/dist/doctor.d.ts +43 -0
- package/dist/doctor.js +425 -0
- package/dist/env-file.d.ts +25 -0
- package/dist/env-file.js +69 -0
- package/dist/env-fill.d.ts +43 -0
- package/dist/env-fill.js +161 -0
- package/dist/env-push.d.ts +37 -0
- package/dist/env-push.js +137 -0
- package/dist/env-registry.d.ts +49 -0
- package/dist/env-registry.js +230 -0
- package/dist/exec.d.ts +14 -0
- package/dist/exec.js +16 -0
- package/dist/fathom-webhook.d.ts +44 -0
- package/dist/fathom-webhook.js +123 -0
- package/dist/fs-utils.d.ts +9 -0
- package/dist/fs-utils.js +34 -0
- package/dist/gcp-service-account-key.d.ts +37 -0
- package/dist/gcp-service-account-key.js +66 -0
- package/dist/gemini-api-key.d.ts +47 -0
- package/dist/gemini-api-key.js +132 -0
- package/dist/git.d.ts +16 -0
- package/dist/git.js +29 -0
- package/dist/github-token.d.ts +24 -0
- package/dist/github-token.js +30 -0
- package/dist/index.d.ts +25 -0
- package/dist/index.js +28 -0
- package/dist/init.d.ts +47 -0
- package/dist/init.js +101 -0
- package/dist/linear-project.d.ts +81 -0
- package/dist/linear-project.js +192 -0
- package/dist/new-client.d.ts +32 -0
- package/dist/new-client.js +89 -0
- package/dist/ping.d.ts +6 -0
- package/dist/ping.js +7 -0
- package/dist/portal-admin.d.ts +133 -0
- package/dist/portal-admin.js +144 -0
- package/dist/sync-templates.d.ts +58 -0
- package/dist/sync-templates.js +109 -0
- package/dist/template-sync.d.ts +55 -0
- package/dist/template-sync.js +90 -0
- package/dist/templates.d.ts +19 -0
- package/dist/templates.js +24 -0
- package/dist/wizard.d.ts +78 -0
- package/dist/wizard.js +213 -0
- package/package.json +30 -0
- package/templates/env.example +60 -0
- package/templates/workspace-root/.claude/settings.json +39 -0
- package/templates/workspace-root/AGENT.md +205 -0
- package/templates/workspace-root/AGENTS.md +11 -0
- package/templates/workspace-root/CLAUDE.md +63 -0
- package/templates/workspace-root/GEMINI.md +24 -0
- package/templates/workspace-root/ONBOARDING.md +164 -0
- package/templates/workspace-root/README.md +67 -0
- package/templates/workspace-root/skills/CLAUDE.md +79 -0
- package/templates/workspace-root/skills/client-deliverables/SKILL.md +131 -0
- package/templates/workspace-root/skills/client-deliverables/references/package-layout.md +103 -0
- package/templates/workspace-root/skills/client-facing-docs/SKILL.md +131 -0
- package/templates/workspace-root/skills/client-facing-docs/references/doc-structures.md +81 -0
- package/templates/workspace-root/skills/client-llm-setup/SKILL.md +97 -0
- package/templates/workspace-root/skills/client-llm-setup/references/guide-template.md +87 -0
- package/templates/workspace-root/skills/client-workflow-diagram/SKILL.md +141 -0
- package/templates/workspace-root/skills/client-workflow-diagram/references/excalidraw-format.md +140 -0
- package/templates/workspace-root/skills/contract-tech-agnostic/SKILL.md +99 -0
- package/templates/workspace-root/skills/contract-tech-agnostic/references/Install_Agent_SOW_Template.docx +0 -0
- package/templates/workspace-root/skills/design-system/SKILL.md +118 -0
- package/templates/workspace-root/skills/design-system/references/tokens.md +139 -0
- package/templates/workspace-root/skills/engagement-spec/SKILL.md +272 -0
- package/templates/workspace-root/skills/engagement-spec/references/spec-templates.md +220 -0
- package/templates/workspace-root/skills/house-voice/SKILL.md +158 -0
- package/templates/workspace-root/skills/house-voice/references/source-copy.md +173 -0
- package/templates/workspace-root/skills/ia-cli/SKILL.md +87 -0
- package/templates/workspace-root/skills/installagent-brand-kit/SKILL.md +112 -0
- package/templates/workspace-root/skills/installagent-brand-kit/assets/installagent-logo.svg +1 -0
- package/templates/workspace-root/skills/installagent-brand-kit/assets/installagent-mark.svg +1 -0
- package/templates/workspace-root/skills/installagent-brand-kit/references/brand-tokens.md +71 -0
- package/templates/workspace-root/skills/installagent-brand-kit/references/branded-pdf.md +66 -0
- package/templates/workspace-root/skills/installagent-brand-kit/references/excalidraw-red-theme.md +73 -0
- package/templates/workspace-root/skills/installagent-brand-kit/scripts/brand.py +175 -0
- package/templates/workspace-root/skills/linear/SKILL.md +186 -0
- package/templates/workspace-root/skills/tech-call-prep/SKILL.md +141 -0
|
@@ -0,0 +1,63 @@
|
|
|
1
|
+
# CLAUDE.md — installagent workspace root
|
|
2
|
+
|
|
3
|
+
Instructions for Claude Code and Cowork working at the root of this workspace.
|
|
4
|
+
|
|
5
|
+
## Read this first
|
|
6
|
+
|
|
7
|
+
1. [`AGENT.md`](./AGENT.md) — canonical conventions and rules for the whole
|
|
8
|
+
workspace. **This is the source of truth**, including the
|
|
9
|
+
"where to start a task" table.
|
|
10
|
+
2. [`ONBOARDING.md`](./ONBOARDING.md) — orientation if you're landing cold.
|
|
11
|
+
3. The `AGENT.md` (or `AGENTS.md`) inside the repo you're working in:
|
|
12
|
+
- [`agents/AGENT.md`](./agents/AGENT.md) — the workflow bank (check it before building any automation; bank back after delivery)
|
|
13
|
+
- [`clients/client-template/AGENT.md`](./clients/client-template/AGENT.md) — the client build repo rules
|
|
14
|
+
- [`clients/client-template/CADENCE.md`](./clients/client-template/CADENCE.md) —
|
|
15
|
+
**how work ships inside any client repo**: the five stages every change
|
|
16
|
+
passes through and the surface matrix that keeps `docs/`,
|
|
17
|
+
`onboarding.html`, and Linear from drifting apart. A copy lives at the
|
|
18
|
+
root of every `clients/client-<slug>/`, synced verbatim — read it before
|
|
19
|
+
changing anything in a client repo, and never write a per-client variant
|
|
20
|
+
- [`team-clients/AGENT.md`](./team-clients/AGENT.md) — the client registry (PII rules!)
|
|
21
|
+
- [`team-infra/AGENT.md`](./team-infra/AGENT.md) — hosting + n8n workflows
|
|
22
|
+
- [`team-web/AGENT.md`](./team-web/AGENT.md) — installagent.com + team hub
|
|
23
|
+
|
|
24
|
+
## How to navigate
|
|
25
|
+
|
|
26
|
+
This root is a folder of **independent git repos** (it is deliberately not a
|
|
27
|
+
git repo itself). To orient on a task:
|
|
28
|
+
|
|
29
|
+
1. **Figure out which repo the task lives in** — use the table in
|
|
30
|
+
`AGENT.md` § *Where to start a task*.
|
|
31
|
+
2. **Client work splits in two:** *who the client is* →
|
|
32
|
+
`team-clients/clients/<slug>/`; *what we're building for them* →
|
|
33
|
+
`clients/client-<slug>/` (start from its `CLIENT.md` and
|
|
34
|
+
`docs/00-delivery-log.md`). Once you're about to *change* something there,
|
|
35
|
+
`CADENCE.md` at that repo's root governs how — including which surfaces
|
|
36
|
+
have to move in the same commit.
|
|
37
|
+
3. **Read the `CLAUDE.md` in the specific directory you're editing.** They're
|
|
38
|
+
navigation aids, not full rule sets — they reference the actual files in
|
|
39
|
+
that directory so you don't have to rediscover the patterns each time.
|
|
40
|
+
|
|
41
|
+
## Behavior
|
|
42
|
+
|
|
43
|
+
- **Refuse pasted secrets.** If the user pastes an API key, token, or
|
|
44
|
+
cookie into the chat, don't use or echo it — tell them it should be
|
|
45
|
+
treated as burned (rotate it), and route them to `ia env fill` (shared
|
|
46
|
+
secrets, from 1Password) or a hand-edit of the root `.env` (personal
|
|
47
|
+
ones). Never `Read`/`cat` the root `.env`; `.claude/settings.json`
|
|
48
|
+
denies it by design.
|
|
49
|
+
- Don't add files at the workspace root without a real reason. The root is
|
|
50
|
+
for cross-cutting things (agentic files, skills, ONBOARDING). Code,
|
|
51
|
+
content, and assets live in the repos.
|
|
52
|
+
- Don't cross-import between repos. They have different audiences, different
|
|
53
|
+
stacks, and different deployment targets. See
|
|
54
|
+
`AGENT.md` § *Cross-cutting golden rules*.
|
|
55
|
+
- The team is currently two people growing to include contractors. When
|
|
56
|
+
writing code or docs, optimize for "a contractor lands here cold next month
|
|
57
|
+
and can be productive within an hour." That's the bar.
|
|
58
|
+
|
|
59
|
+
## Updating conventions
|
|
60
|
+
|
|
61
|
+
If a rule needs to change, edit `AGENT.md` (root or the relevant repo) and
|
|
62
|
+
any skills it references in the same change. Do not edit this file with rule
|
|
63
|
+
content — keep it as a pointer.
|
|
@@ -0,0 +1,24 @@
|
|
|
1
|
+
# GEMINI.md
|
|
2
|
+
|
|
3
|
+
Instructions for the Gemini CLI agent (and other Google-powered agents)
|
|
4
|
+
working at the root of this workspace.
|
|
5
|
+
|
|
6
|
+
## Read these first
|
|
7
|
+
|
|
8
|
+
1. [`AGENT.md`](./AGENT.md) — canonical conventions and rules. **This is the
|
|
9
|
+
source of truth.**
|
|
10
|
+
2. [`README.md`](./README.md) — workspace overview.
|
|
11
|
+
3. The `AGENT.md` (or `AGENTS.md`) inside the repo you're working in:
|
|
12
|
+
- [`clients/client-template/AGENT.md`](./clients/client-template/AGENT.md)
|
|
13
|
+
- [`team-clients/AGENT.md`](./team-clients/AGENT.md)
|
|
14
|
+
- [`team-infra/AGENT.md`](./team-infra/AGENT.md)
|
|
15
|
+
- [`team-web/AGENT.md`](./team-web/AGENT.md)
|
|
16
|
+
|
|
17
|
+
This root is a folder of independent git repos (`clients/client-*`,
|
|
18
|
+
`team-*`); it is deliberately not a git repo itself. Each repo has its own
|
|
19
|
+
conventions — always read the repo-level agentic file before editing inside it.
|
|
20
|
+
|
|
21
|
+
## Updating conventions
|
|
22
|
+
|
|
23
|
+
Edit `AGENT.md` (root or repo) and any affected skills, not this file.
|
|
24
|
+
This file is a pointer.
|
|
@@ -0,0 +1,164 @@
|
|
|
1
|
+
# Install Agent — Onboarding
|
|
2
|
+
|
|
3
|
+
Welcome to the Install Agent workspace. This guide gets you (human or agent)
|
|
4
|
+
productive within one session.
|
|
5
|
+
|
|
6
|
+
## What we build
|
|
7
|
+
|
|
8
|
+
Install Agent productizes AI automations for SMB clients — n8n workflows and
|
|
9
|
+
autonomous agents wrapped with Claude, deployed per-client with
|
|
10
|
+
technician-led setup. The team is currently two people (Angello + one other),
|
|
11
|
+
expanding to contractors.
|
|
12
|
+
|
|
13
|
+
## Setting up from scratch
|
|
14
|
+
|
|
15
|
+
One command clones the whole environment (all team repos, an empty
|
|
16
|
+
`clients/` with `client-template`, these root docs, and `skills/`):
|
|
17
|
+
|
|
18
|
+
```bash
|
|
19
|
+
ia wizard ~/installagent # first-time setup (needs GitHub SSH access)
|
|
20
|
+
ia wizard ~/installagent --update # later: pull every repo + refresh root docs/skills
|
|
21
|
+
```
|
|
22
|
+
|
|
23
|
+
The `ia` CLI lives in `team-tools/` — for the very first run, clone that repo
|
|
24
|
+
alone, `pnpm install && pnpm -r build`, and run `node apps/cli/dist/index.js
|
|
25
|
+
wizard …` (or link it). `--update` is how everyone stays on the same
|
|
26
|
+
environment: fast-forward-only pulls, root docs refreshed only where drifted.
|
|
27
|
+
|
|
28
|
+
The wizard ends with a **numbered next-steps list** — do those in order and
|
|
29
|
+
you're set up. They boil down to:
|
|
30
|
+
|
|
31
|
+
1. `gh auth login` — GitHub auth via one-time OAuth in the browser.
|
|
32
|
+
**No personal access token, no admin-approval wait.** Every `ia` command
|
|
33
|
+
that talks to GitHub uses the gh CLI's token automatically.
|
|
34
|
+
2. `op signin` — the 1Password CLI (`brew install 1password-cli` first if
|
|
35
|
+
needed). Shared team secrets live in the **team-techs** vault.
|
|
36
|
+
3. `ia env fill` — fills the empty vars in the root `.env` straight from
|
|
37
|
+
1Password. No copy-pasting secrets, nothing goes through a chat window.
|
|
38
|
+
4. Two personal values you set by hand in `.env`: your own
|
|
39
|
+
`LINEAR_API_TOKEN` (Linear → Settings → Security & access) and
|
|
40
|
+
`IA_DOCS_SESSION` (session cookie from a logged-in team.installagent.com
|
|
41
|
+
tab).
|
|
42
|
+
5. `ia doctor -w ~/installagent` — repeat until everything required is green.
|
|
43
|
+
|
|
44
|
+
The root `.env` is the ONE env file for the whole workspace: every `ia`
|
|
45
|
+
command auto-loads it, and shell scripts can `set -a; source .env; set +a`.
|
|
46
|
+
It is never overwritten, never committed anywhere; when new vars appear,
|
|
47
|
+
compare against `.env.example` (which `--update` keeps current).
|
|
48
|
+
|
|
49
|
+
## Secrets hygiene (read this even if you skim everything else)
|
|
50
|
+
|
|
51
|
+
- **Never paste a secret value into a Claude (or any AI) chat** — not a PAT,
|
|
52
|
+
not an API key, not a session cookie. A pasted secret lives in the
|
|
53
|
+
conversation history and must be treated as burned: rotate it.
|
|
54
|
+
- Agents working in this workspace must **refuse pasted secrets**: if a
|
|
55
|
+
technician pastes one, say so, don't echo it, and point them at
|
|
56
|
+
`ia env fill` (shared secrets) or "edit `.env` yourself" (personal ones).
|
|
57
|
+
- The right home for a secret is the root `.env` (machine-local, gitignored
|
|
58
|
+
everywhere) with 1Password as the shared source of truth. `ia env fill`
|
|
59
|
+
moves values from the vault to `.env` without a human (or model) ever
|
|
60
|
+
seeing them.
|
|
61
|
+
- The workspace ships `.claude/settings.json` which denies agents read
|
|
62
|
+
access to `.env` — don't work around it; there is never a reason for an
|
|
63
|
+
agent to read that file.
|
|
64
|
+
|
|
65
|
+
## Workspace layout
|
|
66
|
+
|
|
67
|
+
This root folder is **not** a git repo — it holds independent git repos:
|
|
68
|
+
|
|
69
|
+
```
|
|
70
|
+
installagent/
|
|
71
|
+
├── agents/ # the WORKFLOW BANK — reusable n8n templates + plain-language manifests
|
|
72
|
+
├── clients/ # one build repo per client + client-template (the baseline)
|
|
73
|
+
├── team-clients/ # client REGISTRY: who each client is + call history (private, PII allowed)
|
|
74
|
+
├── team-infra/ # AWS hosting (n8n, SFTP), n8n workflow library, automation skills
|
|
75
|
+
├── team-web/ # installagent.com + team hub (Next.js pnpm workspace)
|
|
76
|
+
├── team-design/ # Install Agent design system source
|
|
77
|
+
├── team-branding/ # logo files
|
|
78
|
+
└── skills/ # cross-cutting agent skills (client deliverables pipeline)
|
|
79
|
+
```
|
|
80
|
+
|
|
81
|
+
## Orient yourself in this order
|
|
82
|
+
|
|
83
|
+
1. **Root [`AGENT.md`](./AGENT.md)** — the cross-cutting rules and the
|
|
84
|
+
"where to start a task" table. Five minutes, read it once per session.
|
|
85
|
+
2. **Which client (if any)?** Who they are: `team-clients/clients/<slug>/`
|
|
86
|
+
(profile + call history). What we're building: `clients/client-<slug>/`
|
|
87
|
+
— its `CLIENT.md` maps every external system (Airtable, Drive, n8n,
|
|
88
|
+
1Password), and `docs/00-delivery-log.md` says what has shipped.
|
|
89
|
+
3. **Deal status lives in Airtable** (`Clients - Master`), never in git.
|
|
90
|
+
4. **Read the `AGENT.md` of the repo you're about to touch**, then the
|
|
91
|
+
directory-level `CLAUDE.md` of any folder you edit.
|
|
92
|
+
5. **Open work** is tracked per-repo in `todo/` folders (e.g.
|
|
93
|
+
`team-web/todo/`, `team-infra/todo/`) — check there before starting
|
|
94
|
+
something new.
|
|
95
|
+
6. **Building an automation? Start at the bank.** `agents/` holds every
|
|
96
|
+
reusable n8n template with a plain-language `template.md` manifest
|
|
97
|
+
(what it does, when it fits, what the client must provide) — readable
|
|
98
|
+
even if you never touch JSON. Check its catalog before building from
|
|
99
|
+
scratch, and bank the sanitized result back after delivery: a client
|
|
100
|
+
build isn't done until it's in the bank. See `agents/AGENT.md`.
|
|
101
|
+
|
|
102
|
+
## The non-negotiables
|
|
103
|
+
|
|
104
|
+
1. **No secrets in git** — anywhere, ever. Placeholders only
|
|
105
|
+
(`<KEY_HERE>`, `REPLACE_WITH_*`). Most common foot-gun: real credential
|
|
106
|
+
IDs in n8n workflow JSON.
|
|
107
|
+
2. **No customer data in git.** Exception by design: `team-clients/` holds
|
|
108
|
+
client PII (it's the registry, private, NDA-access) — but never a
|
|
109
|
+
client's *own* customer datasets, and never secrets.
|
|
110
|
+
3. **Client-facing output says Install Agent**, never Rebar.
|
|
111
|
+
4. **pnpm only; Biome only** (no npm/yarn lockfiles, no ESLint/Prettier).
|
|
112
|
+
5. **Run the checks.** Lefthook + `ci/repo-lint.mjs` run pre-commit in
|
|
113
|
+
client repos (`lefthook install` once after cloning). Never `--no-verify`.
|
|
114
|
+
|
|
115
|
+
## AWS
|
|
116
|
+
|
|
117
|
+
All AWS work uses the `installagent` CLI profile:
|
|
118
|
+
|
|
119
|
+
```bash
|
|
120
|
+
aws --profile installagent <cmd> # or export AWS_PROFILE=installagent
|
|
121
|
+
```
|
|
122
|
+
|
|
123
|
+
## Skills
|
|
124
|
+
|
|
125
|
+
Cross-cutting skills live in `skills/` and auto-register in Claude Code.
|
|
126
|
+
|
|
127
|
+
| Skill | When to use |
|
|
128
|
+
|---|---|
|
|
129
|
+
| `client-deliverables` | Orchestrator — full client engagement package |
|
|
130
|
+
| `engagement-spec` | Discovery call → internal docs (overview, tech-spec, onboarding) |
|
|
131
|
+
| `client-facing-docs` | The branded client PDFs |
|
|
132
|
+
| `client-workflow-diagram` | Spec → Excalidraw stage-column diagram (red theme) |
|
|
133
|
+
| `installagent-brand-kit` | Our brand for print/PDF/Excalidraw deliverables |
|
|
134
|
+
| `design-system` | Our brand for screens/UI |
|
|
135
|
+
| `ia-cli` | Route internal ops through the `ia` CLI instead of custom scripts |
|
|
136
|
+
|
|
137
|
+
Repo-local skills: `team-clients/skills/business-research` (lead website →
|
|
138
|
+
registry profile), `team-infra/skills/n8n-workflow.md` (workflow authoring
|
|
139
|
+
rules — read before touching any n8n JSON).
|
|
140
|
+
|
|
141
|
+
## The standard client build repo
|
|
142
|
+
|
|
143
|
+
Created from `clients/client-template/` — flat, fixed, lint-enforced:
|
|
144
|
+
|
|
145
|
+
```
|
|
146
|
+
client-<slug>/
|
|
147
|
+
├── CLIENT.md # context manifest (Airtable, Drive, registry, n8n, 1Password)
|
|
148
|
+
├── onboarding.html # tabbed open-me-first page (Architecture/Client/Proposals/Public)
|
|
149
|
+
├── docs/ # specs + 00-delivery-log.md + HTML diagrams
|
|
150
|
+
├── workflows/ # n8n JSON (credential IDs scrubbed)
|
|
151
|
+
├── data/ # client-sent raw data — gitignored (README tracked)
|
|
152
|
+
├── evidence/ # dated backtest/validation records
|
|
153
|
+
└── web/ design/ agent-<name>/ # optional slots (site, client brand, agent code)
|
|
154
|
+
```
|
|
155
|
+
|
|
156
|
+
No `legal/` (signed docs live in Drive), no `calls/` (registry owns call
|
|
157
|
+
history), no `diagrams/`/`client-facing/` (since 2026-07 — HTML diagrams in
|
|
158
|
+
`docs/`, deliverables via Drive), no nesting, no zips, no status files.
|
|
159
|
+
|
|
160
|
+
## Questions?
|
|
161
|
+
|
|
162
|
+
Ask Angello (angello@installagent.com), or read the client's registry
|
|
163
|
+
profile and build repo delivery log — most context lives in one of those two
|
|
164
|
+
places.
|
|
@@ -0,0 +1,67 @@
|
|
|
1
|
+
# installagent
|
|
2
|
+
|
|
3
|
+
Working folder for Install Agent — the client registry, client build repos,
|
|
4
|
+
infrastructure, web properties, and design system. We productize AI
|
|
5
|
+
automations (n8n workflows + autonomous agents wrapped with Claude) for SMB
|
|
6
|
+
clients and deploy them with technician-led setup sessions.
|
|
7
|
+
|
|
8
|
+
**This root is not a git repo** — it's a folder of independent git repos.
|
|
9
|
+
|
|
10
|
+
## Layout
|
|
11
|
+
|
|
12
|
+
```
|
|
13
|
+
installagent/
|
|
14
|
+
├── clients/ # One build repo per client + client-template (the baseline)
|
|
15
|
+
├── team-clients/ # Client registry (CRM-as-code) — PRIVATE, holds client PII
|
|
16
|
+
├── team-infra/ # AWS CDK hosting (n8n, SFTP), n8n workflow library, skills
|
|
17
|
+
├── team-web/ # installagent.com + team hub (Next.js pnpm workspace)
|
|
18
|
+
├── team-design/ # Install Agent design system source
|
|
19
|
+
├── team-branding/ # Logo files
|
|
20
|
+
├── skills/ # Cross-cutting agent skills
|
|
21
|
+
├── AGENT.md # Canonical rules for all AI coding agents
|
|
22
|
+
└── ONBOARDING.md # New-contractor orientation
|
|
23
|
+
```
|
|
24
|
+
|
|
25
|
+
## The two kinds of client folder
|
|
26
|
+
|
|
27
|
+
| | `team-clients/clients/<slug>/` | `clients/client-<slug>/` |
|
|
28
|
+
|---|---|---|
|
|
29
|
+
| Purpose | **Registry / CRM** — who they are | **Build** — what we make for them |
|
|
30
|
+
| Holds | Profile + call history (PII allowed) | Specs, workflows, agents, deliverables (no PII) |
|
|
31
|
+
| Created | The moment a lead arrives | When build work starts (from `client-template`) |
|
|
32
|
+
|
|
33
|
+
Deal status lives in **Airtable** (`Clients - Master`), never in git.
|
|
34
|
+
|
|
35
|
+
## Quickstarts
|
|
36
|
+
|
|
37
|
+
| Repo | Quickstart |
|
|
38
|
+
|---|---|
|
|
39
|
+
| [`clients/client-template/`](./clients/client-template) | Copy to `clients/client-<slug>/`, fill `CLIENT.md`, `lefthook install` |
|
|
40
|
+
| [`team-web/`](./team-web) | `cd team-web && pnpm install && pnpm dev` |
|
|
41
|
+
| [`team-infra/`](./team-infra) | CDK — see its README; n8n JSON needs no build |
|
|
42
|
+
| [`team-clients/`](./team-clients) | Markdown only — read `CONTRIBUTING.md` before adding a client |
|
|
43
|
+
|
|
44
|
+
## Conventions
|
|
45
|
+
|
|
46
|
+
The canonical rules live in [`AGENT.md`](./AGENT.md). The TL;DR:
|
|
47
|
+
|
|
48
|
+
- **pnpm only**, no npm/yarn lockfiles.
|
|
49
|
+
- **Biome only**, no ESLint/Prettier.
|
|
50
|
+
- **TypeScript strict**, no `any`.
|
|
51
|
+
- **Tailwind utility classes only**.
|
|
52
|
+
- **No secrets anywhere; no customer data** outside the private registry.
|
|
53
|
+
- **Install Agent, never Rebar,** on anything a client sees.
|
|
54
|
+
- **Don't cross-import** between repos.
|
|
55
|
+
|
|
56
|
+
Each repo has its own `AGENT.md`/`AGENTS.md` (canonical) with stack-specific
|
|
57
|
+
rules; many directories have a short `CLAUDE.md` navigation aid.
|
|
58
|
+
|
|
59
|
+
## Who works here
|
|
60
|
+
|
|
61
|
+
Currently two people; expanding to contractors. Start with
|
|
62
|
+
[`ONBOARDING.md`](./ONBOARDING.md).
|
|
63
|
+
|
|
64
|
+
## License
|
|
65
|
+
|
|
66
|
+
Internal. Treat everything as proprietary unless explicitly cleared for
|
|
67
|
+
external sharing.
|
|
@@ -0,0 +1,79 @@
|
|
|
1
|
+
# skills/ — cross-cutting agent skills (technician new-joiner set)
|
|
2
|
+
|
|
3
|
+
Reusable, vendor-neutral how-tos that any coding agent in this workspace reads
|
|
4
|
+
when a task matches the skill's `description`. These live at the workspace
|
|
5
|
+
root because they span repos (a client engagement touches docs, workflows, and
|
|
6
|
+
diagrams across more than one repo) — unlike the repo-local skills in
|
|
7
|
+
`team-clients/skills/` and `team-infra/skills/`.
|
|
8
|
+
|
|
9
|
+
They're mirrored into [`../.claude/skills/`](../.claude/skills/) (via symlink)
|
|
10
|
+
so Claude Code / Cowork auto-register them in the skill menu.
|
|
11
|
+
|
|
12
|
+
## What's here
|
|
13
|
+
|
|
14
|
+
```
|
|
15
|
+
skills/
|
|
16
|
+
├── client-deliverables/ # ORCHESTRATOR — the full client engagement package, checklist + sequencing
|
|
17
|
+
├── engagement-spec/ # discovery call → overview + tech-spec + onboarding docs
|
|
18
|
+
├── tech-call-prep/ # pre-call prep check + per-client call sheet (docs/05-call-sheet.md)
|
|
19
|
+
├── client-llm-setup/ # the client's own Claude workspace → branded leave-behind guide
|
|
20
|
+
├── client-facing-docs/ # the branded client PDFs (exec overview, proposal/SOW, onboarding)
|
|
21
|
+
├── client-workflow-diagram/ # spec → Excalidraw stage-column diagram
|
|
22
|
+
├── installagent-brand-kit/ # our brand for print/PDF/Excalidraw (logo + brand red #9b0506)
|
|
23
|
+
├── design-system/ # the Install Agent Design System (screens/UI)
|
|
24
|
+
├── house-voice/ # house voice for client-facing writing, pulled from installagent.com/get-started
|
|
25
|
+
├── contract-tech-agnostic/ # contracts/SOWs by FUNCTION and OUTCOME, never tech (+ SOW template in references/)
|
|
26
|
+
├── linear/ # Linear issue workflow for client build work (status discipline, sub-issue trees, PR linking)
|
|
27
|
+
└── ia-cli/ # route internal ops through the `ia` CLI (team-tools) — decision map + guardrails
|
|
28
|
+
```
|
|
29
|
+
|
|
30
|
+
Each is `skills/<name>/SKILL.md`, with bulky tables in `references/`.
|
|
31
|
+
|
|
32
|
+
## When to use which
|
|
33
|
+
|
|
34
|
+
| If the task is … | Use |
|
|
35
|
+
|---|---|
|
|
36
|
+
| Kicking off / assembling a full client deliverable package | `client-deliverables` |
|
|
37
|
+
| Writing up a discovery/scoping call into build docs | `engagement-spec` |
|
|
38
|
+
| Prepping for an upcoming client call (sheet + readiness check) | `tech-call-prep` |
|
|
39
|
+
| Setting up a client's own Claude workspace / its leave-behind guide | `client-llm-setup` |
|
|
40
|
+
| Producing the branded client PDFs | `client-facing-docs` |
|
|
41
|
+
| Drawing a workflow/architecture diagram for a client | `client-workflow-diagram` |
|
|
42
|
+
| Branding a print/PDF/Excalidraw deliverable | `installagent-brand-kit` |
|
|
43
|
+
| Building or reviewing any UI / web page / funnel / component | `design-system` |
|
|
44
|
+
| Writing or reviewing any client-facing copy (proposal, email, guide, web/ad copy) | `house-voice` |
|
|
45
|
+
| Drafting/reviewing any contract, SOW, MSA, or proposal | `contract-tech-agnostic` |
|
|
46
|
+
| Working an `INS-xxx` issue / a client's Linear project | `linear` |
|
|
47
|
+
| Any task an `ia` command covers (client context, intake, agent bank, secrets, setup/doctor) | `ia-cli` |
|
|
48
|
+
|
|
49
|
+
`design-system` is the **screen** layer (dark, signal-red, Hanken Grotesk); it
|
|
50
|
+
mirrors the code source of truth `team-web/packages/ui`.
|
|
51
|
+
`installagent-brand-kit` is the **print/deliverable** layer for PDFs and
|
|
52
|
+
Excalidraw on white. They share one brand and must agree. `house-voice` is
|
|
53
|
+
the **words** layer that applies across both — what client-facing sentences
|
|
54
|
+
sound like, regardless of surface.
|
|
55
|
+
|
|
56
|
+
These complement the repo-local skills they cross-reference:
|
|
57
|
+
`business-research` (team-clients) and `n8n-workflow` + `automation-design`
|
|
58
|
+
(team-infra).
|
|
59
|
+
|
|
60
|
+
## Convention
|
|
61
|
+
|
|
62
|
+
- One skill per directory: `skills/<name>/SKILL.md` with YAML frontmatter
|
|
63
|
+
(`name`, `description`) — same format as `team-clients/skills/`.
|
|
64
|
+
- The `description` is the trigger: say what the skill does **and** when to use
|
|
65
|
+
it, concretely. Put all "when to use" cues there, not in the body.
|
|
66
|
+
- Keep `SKILL.md` under ~500 lines; push long tables/detail into `references/`.
|
|
67
|
+
- Keep skills sanitized — no secrets, live credentials, internal hostnames,
|
|
68
|
+
account ids, or PII. Use placeholders and point at internal files for the
|
|
69
|
+
real values.
|
|
70
|
+
- If a rule in a skill conflicts with [`../AGENT.md`](../AGENT.md), **AGENT.md
|
|
71
|
+
wins** — update the skill in the same change.
|
|
72
|
+
|
|
73
|
+
## Adding a skill
|
|
74
|
+
|
|
75
|
+
1. Create `skills/<name>/SKILL.md` (frontmatter + body). Add `references/` for
|
|
76
|
+
bulky tables.
|
|
77
|
+
2. Mirror it: `cd ../.claude/skills && ln -s ../../skills/<name> <name>`.
|
|
78
|
+
3. Add it to the table above. Don't put workspace-wide rules here — those live
|
|
79
|
+
in `AGENT.md`. Skills are how-tos.
|
|
@@ -0,0 +1,131 @@
|
|
|
1
|
+
---
|
|
2
|
+
name: client-deliverables
|
|
3
|
+
description: >-
|
|
4
|
+
The standard deliverable package Install Agent produces for EVERY client
|
|
5
|
+
engagement. Three things, always: (1) an in-depth internal HTML overview
|
|
6
|
+
(CEO/pricing brief, architecture, tech spec, roadmap/intake), (2) three
|
|
7
|
+
client-facing PDFs (executive overview, proposal & SOW, onboarding
|
|
8
|
+
checklist), and (3) workflow/architecture diagrams. Use this WHENEVER
|
|
9
|
+
standing up a new client engagement, asked "set up the deliverables for
|
|
10
|
+
[client]", "build the client folder", "what do we owe this client", "package
|
|
11
|
+
this up", or working in a client-[slug]/ folder needing the complete set.
|
|
12
|
+
Reach for it even without the word "skill" — any time the goal is the full
|
|
13
|
+
artifact set an engagement should produce. It is an ORCHESTRATOR — composes
|
|
14
|
+
engagement-spec, client-facing-docs, client-workflow-diagram, and
|
|
15
|
+
installagent-brand-kit — and defines the flat client-[slug]/ folder layout
|
|
16
|
+
(per client-template) the new-lead workflow will eventually enforce.
|
|
17
|
+
---
|
|
18
|
+
|
|
19
|
+
# Client deliverables (the standard package)
|
|
20
|
+
|
|
21
|
+
Every client engagement produces the **same three kinds of artifact**, so any
|
|
22
|
+
teammate (or a contractor landing cold) finds the same things in the same place:
|
|
23
|
+
|
|
24
|
+
1. **In-depth internal HTML overview** — one self-contained, tabbed `.html`
|
|
25
|
+
that consolidates the full internal picture: CEO/pricing brief · architecture
|
|
26
|
+
· technical spec · roadmap & intake. This is the *deep* version (names tools,
|
|
27
|
+
shows the decision model), more thorough than anything the client sees. Lives
|
|
28
|
+
in `docs/`. Exemplars: `client-kingdom/docs/kingdom-workflow-overview.html`,
|
|
29
|
+
`client-carvision/pricing-agent/client-facing/carvision-pricing-overview.html`.
|
|
30
|
+
2. **The three client-facing PDFs** — executive overview, proposal & SOW,
|
|
31
|
+
onboarding checklist. Built by **`client-facing-docs`** (branded, decision-maker
|
|
32
|
+
voice).
|
|
33
|
+
3. **Workflow / architecture diagrams** — the how-it-works picture and any
|
|
34
|
+
mockups, authored as self-contained HTML in `docs/` (2026-07 baseline; no
|
|
35
|
+
`.excalidraw` in client repos — Excalidraw only for pitch-deck images).
|
|
36
|
+
Built by **`client-workflow-diagram`** in the brand's red theme
|
|
37
|
+
(**`installagent-brand-kit`**).
|
|
38
|
+
|
|
39
|
+
Plus **`evidence/`** — required scaffolding (per `client-template`) for dated
|
|
40
|
+
backtest/validation records. Most engagements have nothing to put there until
|
|
41
|
+
there's a decision/prediction component worth validating; it's fine for it to
|
|
42
|
+
hold only its `README.md` until then, but the directory itself must exist
|
|
43
|
+
(`ci/repo-lint.mjs` checks for it).
|
|
44
|
+
|
|
45
|
+
This skill is the **checklist + orchestrator** for producing all three; it does
|
|
46
|
+
not replace the skills that make each piece — it sequences them and owns the
|
|
47
|
+
folder layout. Full layout + build order:
|
|
48
|
+
[`references/package-layout.md`](references/package-layout.md).
|
|
49
|
+
|
|
50
|
+
## The folder (one per client)
|
|
51
|
+
|
|
52
|
+
Flat, single-project layout — created from `client-template`, enforced per-repo
|
|
53
|
+
by its `ci/repo-lint.mjs`:
|
|
54
|
+
|
|
55
|
+
```
|
|
56
|
+
client-<slug>/
|
|
57
|
+
├── AGENT.md / CLAUDE.md / CLIENT.md
|
|
58
|
+
├── onboarding.html # tabbed open-me-first page: Architecture/Client/Proposals/Public
|
|
59
|
+
├── docs/ # internal markdown (engagement-spec): 01-overview, 02-tech-spec,
|
|
60
|
+
│ # 03-client-onboarding, ① the in-depth internal HTML (tabbed),
|
|
61
|
+
│ # plus ③ diagrams as self-contained HTML (<system>-<audience>-diagram.html)
|
|
62
|
+
├── workflows/ # n8n JSON exports, credential IDs scrubbed
|
|
63
|
+
├── data/ # client-sent raw data — gitignored (README tracked)
|
|
64
|
+
├── evidence/ # dated backtest/validation records — required, README-only
|
|
65
|
+
│ # until there's a decision/prediction component to validate
|
|
66
|
+
└── web/ # OPTIONAL — pre-populated Next.js scaffold, only if the
|
|
67
|
+
# engagement includes a website build
|
|
68
|
+
```
|
|
69
|
+
|
|
70
|
+
② the three branded PDFs (client-facing-docs) are delivered via the client's
|
|
71
|
+
Google Drive folder — there is no `client-facing/` directory anymore, no
|
|
72
|
+
`diagrams/`, and there was never a `legal/` (signed SOW/MSA live in Drive).
|
|
73
|
+
|
|
74
|
+
Don't nest a `<project>/` subfolder — one client is one build tree. (CarVision's
|
|
75
|
+
`pricing-agent/` predates this convention and keeps PDFs/diagrams nested; new
|
|
76
|
+
clients follow the flat layout above.)
|
|
77
|
+
|
|
78
|
+
## Build order (compose, don't reinvent)
|
|
79
|
+
|
|
80
|
+
1. **Understand the client** — `business-research` → a registry profile.
|
|
81
|
+
2. **Scaffold the engagement** — `client-project-kickoff` (the client-folder
|
|
82
|
+
cadence): clarifying questions → research → plan → folder + CLAUDE.md.
|
|
83
|
+
3. **Write the internal docs** — `engagement-spec` → `docs/` markdown
|
|
84
|
+
(overview / tech-spec / onboarding). The source of truth for content.
|
|
85
|
+
4. **Consolidate the in-depth internal HTML** — turn the briefs + tech spec +
|
|
86
|
+
roadmap into one tabbed `.html` (see the layout reference). Internal: names
|
|
87
|
+
tools, shows the model.
|
|
88
|
+
5. **Produce the client PDFs** — `client-facing-docs` (branded; the commercial
|
|
89
|
+
model incl. capped API-credit pass-through).
|
|
90
|
+
6. **Draw the diagrams** — `client-workflow-diagram` + `installagent-brand-kit`
|
|
91
|
+
(red theme); export images for the deck/overview.
|
|
92
|
+
7. **Leave `evidence/` scaffolded** — the README-only convention if there's
|
|
93
|
+
nothing to validate yet; a dated record once there is.
|
|
94
|
+
|
|
95
|
+
## Quality rules
|
|
96
|
+
|
|
97
|
+
- **All three artifact types, every client.** A folder missing the internal
|
|
98
|
+
HTML, the PDF set, or a diagram is incomplete.
|
|
99
|
+
- **`evidence/` and (if scoped) `web/` exist.** `evidence/` is required
|
|
100
|
+
scaffolding even when README-only; `web/` is optional and only present if
|
|
101
|
+
the engagement includes a website build — delete it entirely otherwise.
|
|
102
|
+
- **Internal vs client split holds throughout:** the internal HTML and `docs/`
|
|
103
|
+
name tools; the PDFs stay tech-agnostic (Exhibit A aside). See
|
|
104
|
+
`engagement-spec` / `contract-tech-agnostic`.
|
|
105
|
+
- **One brand:** every visual (HTML, PDFs, diagrams) uses
|
|
106
|
+
`installagent-brand-kit` (official logo + brand red).
|
|
107
|
+
- **No secrets / customer data / real VINs** in committed artifacts (root
|
|
108
|
+
`AGENT.md`); use redacted/sample data.
|
|
109
|
+
|
|
110
|
+
## Enforcement
|
|
111
|
+
|
|
112
|
+
`client-template`'s `ci/repo-lint.mjs` enforces the skeleton (`docs/`,
|
|
113
|
+
`workflows/`, `data/`, `evidence/` and `onboarding.html` must exist; the
|
|
114
|
+
`.gitignore` must keep `data/` untracked) on every
|
|
115
|
+
push for repos created from the template. It does not check for the in-depth
|
|
116
|
+
internal HTML, the PDFs themselves, or a diagram's contents — that's still this
|
|
117
|
+
skill's job. Full automatic scaffolding on lead conversion (the new-lead /
|
|
118
|
+
new-client n8n workflow, `team-n8n-workflows`) is recorded as a project-level
|
|
119
|
+
intention in root `AGENT.md` but **not built yet**; until then, use
|
|
120
|
+
`client-template` to start a new client repo and this skill as the manual
|
|
121
|
+
checklist for what goes in it.
|
|
122
|
+
|
|
123
|
+
## Before you finish — checklist
|
|
124
|
+
|
|
125
|
+
- [ ] In-depth internal HTML exists (tabbed: brief · architecture · spec · roadmap), in `docs/`.
|
|
126
|
+
- [ ] The three client PDFs exist (via `client-facing-docs`), on-brand.
|
|
127
|
+
- [ ] At least one workflow/architecture diagram exists (editable + exported).
|
|
128
|
+
- [ ] `evidence/` exists (README-only is fine); `web/` present only if a website was scoped.
|
|
129
|
+
- [ ] Folder matches the flat layout (no nested `<project>/`); `CLAUDE.md` orients the next person.
|
|
130
|
+
- [ ] Internal names tools; client-facing stays tech-agnostic; one brand throughout.
|
|
131
|
+
- [ ] No secrets/customer data; branding is Install Agent.
|
|
@@ -0,0 +1,103 @@
|
|
|
1
|
+
# Package layout + the in-depth internal HTML
|
|
2
|
+
|
|
3
|
+
Detail for the standard per-client deliverable package. Pairs with the SKILL.md.
|
|
4
|
+
|
|
5
|
+
## Folder layout
|
|
6
|
+
|
|
7
|
+
Flat, single-project layout — created from `client-template`, enforced per-repo
|
|
8
|
+
by its `ci/repo-lint.mjs`. Don't nest a `<project>/` subfolder; if a client ever
|
|
9
|
+
grows a second genuinely distinct project, raise it and decide the shape
|
|
10
|
+
deliberately rather than guessing.
|
|
11
|
+
|
|
12
|
+
```
|
|
13
|
+
client-<slug>/
|
|
14
|
+
├── AGENT.md # canonical agent instructions
|
|
15
|
+
├── CLAUDE.md # pointer + navigation aid
|
|
16
|
+
├── CLIENT.md # context manifest (Airtable, Drive, registry, no PII)
|
|
17
|
+
├── docs/ # engagement-spec output (internal markdown)
|
|
18
|
+
│ ├── 01-overview.md # client-facing prose, tech-agnostic
|
|
19
|
+
│ ├── 02-tech-spec.md # internal: names tools, the decision model
|
|
20
|
+
│ ├── 03-client-onboarding.md # what we need from the client
|
|
21
|
+
│ └── <system>-overview.html # ① in-depth INTERNAL html (tabbed), if built
|
|
22
|
+
├── onboarding.html # tabbed open-me-first page (Architecture/Client/
|
|
23
|
+
│ # Proposals/Public — Public is client-presentable)
|
|
24
|
+
├── workflows/ # n8n JSON exports, credential IDs scrubbed
|
|
25
|
+
├── data/ # client-sent raw data — gitignored (README tracked)
|
|
26
|
+
├── evidence/ # ④ dated backtest/validation records — required,
|
|
27
|
+
│ └── README.md # README-only until there's something to validate
|
|
28
|
+
└── web/ # OPTIONAL — pre-populated Next.js marketing site,
|
|
29
|
+
# only if the engagement includes a website build
|
|
30
|
+
```
|
|
31
|
+
|
|
32
|
+
③ diagrams are self-contained HTML in `docs/` (`<system>-<audience>-diagram.html`;
|
|
33
|
+
no `.excalidraw` in client repos since 2026-07). ② the three branded client PDFs
|
|
34
|
+
are delivered via the client's Google Drive folder — `client-facing/` and
|
|
35
|
+
`diagrams/` directories no longer exist, and `legal/` never belonged in git
|
|
36
|
+
(signed SOW/MSA live in Drive).
|
|
37
|
+
|
|
38
|
+
Keep internal artifacts (`docs/`, the HTML) and client-facing artifacts (the
|
|
39
|
+
Drive-delivered PDFs, the Public tab of `onboarding.html`) distinguishable —
|
|
40
|
+
those are the only things a client sees.
|
|
41
|
+
|
|
42
|
+
(CarVision's `pricing-agent/` predates this flat convention and nests its HTML,
|
|
43
|
+
PDFs, and diagrams under a project subfolder — it's a legacy exemplar for
|
|
44
|
+
content/structure, not for folder layout; follow the flat layout above for new
|
|
45
|
+
clients.)
|
|
46
|
+
|
|
47
|
+
## ① The in-depth internal HTML
|
|
48
|
+
|
|
49
|
+
A single self-contained `.html` (inline CSS/JS, no build step) with **tabs**
|
|
50
|
+
that consolidate the internal picture into one navigable file, living in
|
|
51
|
+
`docs/`. Standard tabs, modeled on
|
|
52
|
+
`client-kingdom/docs/kingdom-workflow-overview.html` and
|
|
53
|
+
`client-carvision/pricing-agent/client-facing/carvision-pricing-overview.html`
|
|
54
|
+
(CarVision predates the current `docs/` placement):
|
|
55
|
+
|
|
56
|
+
| Tab | Contents |
|
|
57
|
+
|---|---|
|
|
58
|
+
| **CEO Brief & Pricing** | The business case + commercial rationale (problem, what we're building, why it's worth it, recommended pricing incl. the capped API-credit pass-through, what to set expectations on). |
|
|
59
|
+
| **Architecture** | The how-it-works diagram + a short narration. Embed the exported diagram image. |
|
|
60
|
+
| **Technical Spec** | The full internal spec: goal, current-state process, systems & data sources, the decision model, architecture, phasing, guardrails, risks, "what we need", effort→pricing rationale. Names tools. |
|
|
61
|
+
| **Roadmap & Intake** | The phased roadmap + the intake checklist (internal view). |
|
|
62
|
+
|
|
63
|
+
It's the deep, single-link reference for the team and contractors — more
|
|
64
|
+
thorough than the three client PDFs, which are distilled from it. Branding via
|
|
65
|
+
`installagent-brand-kit` (official logo + brand red); it can stay light on
|
|
66
|
+
chrome since it's internal, but keep the palette consistent.
|
|
67
|
+
|
|
68
|
+
Build it from the `engagement-spec` markdown (the briefs + tech spec + roadmap)
|
|
69
|
+
— the HTML is a *view* of those, not a separate source of truth. When the spec
|
|
70
|
+
changes, regenerate the HTML in the same pass.
|
|
71
|
+
|
|
72
|
+
## ② Client PDFs
|
|
73
|
+
|
|
74
|
+
Built by **`client-facing-docs`** → `references/doc-structures.md`. Branded,
|
|
75
|
+
decision-maker voice, the full commercial model.
|
|
76
|
+
|
|
77
|
+
## ③ Diagrams
|
|
78
|
+
|
|
79
|
+
Built by **`client-workflow-diagram`** (layout: left-to-right stage columns,
|
|
80
|
+
emoji node names; client-facing + internal flavors) skinned with
|
|
81
|
+
**`installagent-brand-kit`** (red theme). Commit the `.excalidraw`; export an
|
|
82
|
+
image for the deck and to embed in the Architecture tab of the internal HTML.
|
|
83
|
+
|
|
84
|
+
## ④ Evidence
|
|
85
|
+
|
|
86
|
+
`evidence/` holds dated, self-contained `.md` records of backtests/validation
|
|
87
|
+
runs against the client's real data — proof the automation does what we said
|
|
88
|
+
it would, meant to be exported to the client's Drive. Required scaffolding on
|
|
89
|
+
every client repo (per `client-template`); most engagements have nothing to
|
|
90
|
+
put there until there's a decision/prediction component (pricing, scoring,
|
|
91
|
+
categorization, routing) worth backtesting, so it's fine to hold only its
|
|
92
|
+
`README.md` until then. See `evidence/README.md`'s convention for filenames
|
|
93
|
+
and format — one file per run, no raw customer PII, aggregate results only.
|
|
94
|
+
|
|
95
|
+
## Composing the skills (order)
|
|
96
|
+
|
|
97
|
+
`business-research` → `client-project-kickoff` → `engagement-spec` →
|
|
98
|
+
in-depth internal HTML → `client-facing-docs` → `client-workflow-diagram` +
|
|
99
|
+
`installagent-brand-kit` → scaffold `evidence/` (populate once there's a
|
|
100
|
+
validation run to record).
|
|
101
|
+
|
|
102
|
+
Each step has its own skill that owns the details; this package just guarantees
|
|
103
|
+
the artifact types land, in the right folder, on-brand.
|