@installagent/sdk 0.14.0 → 0.16.0

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.
@@ -28,6 +28,9 @@ AIRTABLE_API_TOKEN=
28
28
  # read-only Airtable PAT (data.records:read on Clients - Master) — `ia env fill`
29
29
  GOOGLE_SERVICE_ACCOUNT_KEY=
30
30
  # the Drive service-account key JSON, single line — filled by `ia env fill`
31
+ IA_PREP_CALENDAR_ID=
32
+ # optional — the calendar id for `ia tech-call prep`. Share the calendar read-only
33
+ # with the Drive service account first; unset = the invite check is not run.
31
34
  FATHOM_API_KEY=
32
35
  # Fathom dashboard → Settings → API Access — filled by `ia env fill`
33
36
  N8N_API_URL=
@@ -45,7 +45,8 @@ installagent/
45
45
  | Task | Go to | Read first |
46
46
  |---|---|---|
47
47
  | Understand who a client/lead is, prep a call | `team-clients/clients/<slug>/` | `team-clients/AGENT.md` |
48
- | Prep a tech discovery / scoping call (call agenda, proposals ready) | `clients/client-<slug>/` via the `tech-call-prep` skill | team hub → Techs → Discovery call |
48
+ | Prep a tech discovery / scoping call (call agenda, proposals ready) | `clients/client-<slug>/` via the `tech-call-prep` skill: `ia tech-call verify`, `agenda`, `prep` | team hub → Techs → Discovery call |
49
+ | Update a client's docs after the tech discovery call | the bot's branch `bot/tech-call-update-<date>` (`ia tech-call update-pass`); read its `evidence/` file, then merge | the `engagement-spec` skill § *The two passes now have commands* |
49
50
  | Build/edit anything *for* a client (spec, workflow, diagram, PDF, site, agent) | `clients/client-<slug>/` | that repo's `CADENCE.md` (how to ship it) + `AGENT.md` + `CLIENT.md` |
50
51
  | Ship a change into a client repo without causing drift (feature, fix, scope change) | `clients/client-<slug>/CADENCE.md` | it *is* the read — five stages + the surface matrix |
51
52
  | Answer "what did we ship for X / does it work?" | `clients/client-<slug>/docs/00-delivery-log.md` + `evidence/` | same |
@@ -115,6 +115,33 @@ This skill runs **twice** per engagement, against two different calls in
115
115
  - Move resolved items out of `02-tech-spec.md`'s open-questions list
116
116
  rather than deleting the section.
117
117
 
118
+ ### The two passes now have commands
119
+
120
+ Both passes run by themselves. This skill is still the rule set: the commands
121
+ follow it, and you use it to judge and to correct what they wrote.
122
+
123
+ - **Pass 1** is the lead-close kickoff's *Draft the engagement docs* step
124
+ (team hub → Sales → Lead Close Kickoff). With no sales-call note and no
125
+ recording it writes **no docs**: the step shows **Waiting for the call note**
126
+ and the repo page shows a WAITING FOR CALL banner. It runs by itself when
127
+ the note is merged in `team-clients`. Never fill the docs by hand from the
128
+ intake profile to get past the banner: that is how a client got a spec for a
129
+ build they never asked for.
130
+ - **After pass 1**, `ia tech-call verify <slug> --write` checks every claim in
131
+ docs `01`–`04` against the transcripts and the client's Slack channel. A line
132
+ with an `UNVERIFIED` or `CONTRADICTED` mark is a line no call says: fix it or
133
+ confirm it, then run the command again.
134
+ - **Pass 2** is `ia tech-call update-pass <slug>`. When the tech-discovery
135
+ note lands, it runs this skill's update pass with Claude Code in a scratch
136
+ clone and pushes the branch `bot/tech-call-update-<date>`. The technician
137
+ gets one Slack DM, reads `evidence/<date>-tech-call-update.md` and the diff,
138
+ and merges. It keeps lines a person wrote unless the call contradicts them,
139
+ and it only DRAFTS Linear issues: you create and close them (ask first,
140
+ Linear is shared).
141
+
142
+ Run a pass by hand with this skill when a command cannot (no transcript, a
143
+ call that was not recorded), and say so in the delivery-log entry.
144
+
118
145
  ## Workflow
119
146
 
120
147
  ### 1. Extract the five scoping answers
@@ -54,6 +54,11 @@ secrets).
54
54
  | Set up a new email-triage client's record / "the record node is wired, now what?" | `ia portal create-agent …`, bind the two credentials in the n8n UI, then `ia triage enable <slug>` BEFORE the first publish — checks the live workflow's record wiring (read-only, node names + pass/fail only), marks the agent (`ia portal set-agent-triage`), backfills an old sheet if one exists. `--dry-run` checks only; exit 1 on a failed check, with the fix |
55
55
  | "Record did not reach the portal" alert in `#team-techs` | copy that execution's `🧺 Batch Record Rows` output into a file → `ia triage backfill <slug> --rows <file>` (idempotent; refuses rows from another client's workflow). `ia triage backfill <slug> [--sheet\|--csv]` copies an old record sheet in; a re-run that says 0 new is the parity check |
56
56
  | Weekly triage summary into the client's channel | `ia portal set-org-slack-channel --slug <slug> --channel <C…>` (`none` clears it) — the digest only posts for an org with a channel |
57
+ | "Do the docs say what the client said?" / check a spec against the recording | `ia tech-call verify <slug>` — every claim in docs `01`–`04` against the call transcripts and the client's Slack channel; `--write` marks what no call says and writes `evidence/<date>-transcript-check.md`; `--ref <git-ref>` checks a commit without touching the working tree |
58
+ | Generate or refresh the call agenda (there is ONE) | `ia tech-call agenda <slug> --call-date <date>` — draft only; `--publish` updates the one Drive Doc in place, `--write` writes `docs/05-call-agenda.md`. Never hand-make a second agenda Doc |
59
+ | "Am I ready for the [client] call?" | `ia tech-call prep <slug>` — repo, docs vs the latest call, one agenda Doc, the invite, the vault, the portal login, Linear; `--write` fills the Prep gaps block. NOT CHECKED is not a pass |
60
+ | Update the docs after the tech discovery call | `ia tech-call update-pass <slug>` — dry run in a scratch clone; `--push --notify` pushes `bot/tech-call-update-<date>` and DMs the technician. A person merges. CLI-only |
61
+ | A call note landed but the kickoff still waits / replay the event | `ia tech-call note-landed <slug> [--file <date>.md]` |
57
62
  | Mint a GCP SA key / Gemini key / Fathom webhook | `ia secrets mint <kind> …` |
58
63
  | Portal org/agent admin (create org, passwords, run sync) | `ia portal <cmd>` — CLI-only, never via MCP |
59
64
  | A build went live / "what would this build cost at market rate?" | add its row to `CLIENT.md` `## Builds`, then `ia build-estimate --build <slug>` in the client repo: market hours (low–high) + a one-line basis, from one Gemini call on the build's spec. Hours only — never a rate or a dollar figure, in git, Linear or Slack. Review it, then `--write` sets the row's `est_hours`; commit it with the row. The portal multiplies by its one rate (admin Clients page); `ia portal sync-builds --apply` (account owner) carries it over. `--force` re-estimates a row; `--spec <file>` points it at the right spec |
@@ -90,6 +90,19 @@ line:` is internal: the client never sees it. The portal never shows an issue
90
90
  title or any other part of the description, and drops a line with `$` or a
91
91
  ticket id.
92
92
 
93
+ **Something the client must do or answer** goes on their portal To-do tab
94
+ from Linear. Give the issue the `client-todo` label and one line in its
95
+ description: `Client to-do: <what the client must do>` or `Client question:
96
+ <the question>` (160 characters at most; no ticket number, no dollar amount, no
97
+ email address; the client never sees that a ticket is behind the item). The
98
+ portal follows the issue: the client's tick moves it to **In Review** with a
99
+ comment; move it to **Done** to confirm the item for the client, or back to In
100
+ Progress to open the item again; cancel the issue or remove the label to take
101
+ the item off the tab. Access the client still owes needs no issue: a MISSING
102
+ or REQUESTED row in `docs/04-access-checklist.md` §1 is an item, and HAVE
103
+ confirms it. `ia portal sync-todo --slug <slug>` runs the sync now; the portal
104
+ also runs it when the client opens the tab and once a day.
105
+
93
106
  **A new client ask** is a new issue in the client's project: its `Client
94
107
  line:`, the `client-ask` label, and a priority. The account owner places it.
95
108
 
@@ -7,9 +7,9 @@ description: >-
7
7
  close checklist). Reads the registry (team-clients/clients/[slug]/) and the
8
8
  build repo (clients/client-[slug]/ docs 01–04, onboarding.html), checks the
9
9
  come-prepared checklist (repo scaffolded, engagement docs drafted, proposals
10
- loaded), pulls in the auto-drafted agenda Doc from the client's Drive folder
11
- when one exists, and writes docs/05-call-agenda.md plus a Google Doc copy in
12
- that folder. Use this WHENEVER a discovery/scoping call is coming up, or
10
+ loaded) with `ia tech-call prep`, and generates the ONE agenda with
11
+ `ia tech-call agenda` (docs/05-call-agenda.md plus the one Google Doc in the
12
+ client's Drive folder, updated in place). Use this WHENEVER a discovery/scoping call is coming up, or
13
13
  someone says "prep me for the call with [client]", "generate the call agenda
14
14
  / call sheet", "am I ready for the [client] call?", or you are in a
15
15
  client-[slug]/ repo before a call. The static call script lives on the team
@@ -30,6 +30,24 @@ transcript evidence shows a written agenda that stayed closed while the two
30
30
  blocks that most needed it (access audit, decisions) got dropped. Print it
31
31
  or keep it on a second screen.
32
32
 
33
+ ## The commands do the lookups; you do the judgment
34
+
35
+ Three `ia` commands (the `ia-cli` skill) replace the hand work of this skill.
36
+ Run them in this order, then do what only a person can do:
37
+
38
+ | Command | What it does |
39
+ |---|---|
40
+ | `ia tech-call verify <slug> --write` | Checks every claim in docs `01`–`04` against the call transcripts and the client's Slack channel. Marks what no call says (`UNVERIFIED`, `CONTRADICTED`) and writes `evidence/<date>-transcript-check.md`, with the promises we made on a call. |
41
+ | `ia tech-call agenda <slug> --call-date <date> --publish --write` | Generates the ONE agenda (the format in step 3) from the notes and transcripts, docs `01`–`04` and the industry playbook. Updates the one Drive Doc in place and writes `docs/05-call-agenda.md`. |
42
+ | `ia tech-call prep <slug> --write` | Runs the machine checks of step 2 and writes the Prep gaps block of `docs/05`. A check it could not run says NOT CHECKED: do that one by hand. |
43
+
44
+ There is ONE agenda. The lead-close kickoff drafts it with the same code, and
45
+ a second run updates the same Drive Doc (`<call date>-tech-discovery-agenda`).
46
+ Never make a second agenda Doc. `ia tech-call prep` lists stale ones: delete them.
47
+
48
+ The steps below say what a good agenda holds and what you still check by
49
+ hand. Read the generated agenda against them; do not write it from a blank page.
50
+
33
51
  ## Inputs
34
52
 
35
53
  - **Client slug** (required). Registry: `team-clients/clients/<slug>/`.
@@ -73,15 +91,15 @@ name mis-transcriptions and discrepancies live), every file in registry
73
91
  `onboarding.html` (Proposals tab). For a monthly scoping
74
92
  call, add `docs/00-delivery-log.md` and `evidence/`.
75
93
 
76
- **Check for the auto-drafted agenda.** The portal's Lead Close Kickoff
77
- (sales runs it the moment a deal closes — team hub → Sales → Lead Close
78
- Kickoff) drops a `<date>-tech-discovery-agenda` Google Doc at the **root
79
- of the client's Drive folder**, drafted from the merged registry call
80
- notes. If one exists, use it as the first draft — but verify every fact in it
81
- against the registry and the sales-call notes before keeping it. It is a
82
- draft from a transcript, not a source of truth; transcripts mangle names
83
- and the model can mis-rank priorities. Never run a call from an unverified
84
- auto-draft.
94
+ **The agenda already exists.** The portal's Lead Close Kickoff (sales runs
95
+ it when a deal closes — team hub → Sales → Lead Close Kickoff) drafts the
96
+ `<date>-tech-discovery-agenda` Google Doc at the **root of the client's Drive
97
+ folder**, and `ia tech-call agenda` regenerates it. It is a draft from a
98
+ transcript, not a source of truth: transcripts mangle names and the model can
99
+ mis-rank priorities. Verify every fact in it against the registry and the
100
+ recording. Never run a call from an agenda nobody verified. If the kickoff's
101
+ agenda and docs steps show **Waiting for the call note**, the sales-call note
102
+ is not merged yet: merge it, and they run by themselves.
85
103
 
86
104
  Then derive the client's **vertical** from registry `PROFILE.md` §
87
105
  "Industry & business model" (primary industry / NAICS) and look for a
@@ -95,7 +113,11 @@ on the agenda as a prep-gap line ("no industry playbook for
95
113
 
96
114
  ### 2. Verify the come-prepared checklist
97
115
 
98
- Check, honestly, against the build repo:
116
+ `ia tech-call prep <slug> --write` checks the rows a machine can check (the
117
+ repo, the docs against the latest call, the one agenda Doc, the calendar
118
+ invite, the vault, the portal login, the Linear issues) and writes the result
119
+ into the Prep gaps block. You check what it cannot — the last row here, and
120
+ anything it reports as NOT CHECKED:
99
121
 
100
122
  | Check | Pass looks like |
101
123
  | --- | --- |
@@ -110,7 +132,14 @@ with the fix (usually: run `engagement-spec`, or fill the Proposals tab).
110
132
  If the repo doesn't exist at all, stop and say so — scaffolding comes
111
133
  first (`client-template`), not a call agenda.
112
134
 
113
- ### 3. Write `docs/05-call-agenda.md`
135
+ ### 3. Generate `docs/05-call-agenda.md`
136
+
137
+ `ia tech-call agenda <slug> --call-date <date> --publish --write` writes it in
138
+ this structure (the model writes the nine sections from Read-back to Close
139
+ checklist; the command writes the title, the notes under it and the Prep gaps
140
+ block). To correct a fact, fix the source — docs `01`–`04` or the registry —
141
+ and run the command again. For a line the command cannot know, edit
142
+ `docs/05` and paste the same line into the Drive Doc, so the two stay equal.
114
143
 
115
144
  One file in the build repo, overwritten per call (it's a prep artifact,
116
145
  not a record — records are registry call notes). The line under its title
@@ -188,11 +217,11 @@ access I need. At the end we book the follow-ups."
188
217
 
189
218
  ### 4. Copy the agenda to the client's Drive
190
219
 
191
- Place a copy of the finished agenda at the **root of the client's Drive
192
- folder**, named `<YYYY-MM-DD>-tech-discovery-agenda` (as a Google Doc, so
193
- the technician can read it from any device mid-call). If the kickoff's
194
- auto-drafted Doc from step 1 exists, update **that Doc in place** rather
195
- than creating a second one — one agenda per call, everywhere. The Drive
220
+ `--publish` does this: the agenda is at the **root of the client's Drive
221
+ folder**, named `<YYYY-MM-DD>-tech-discovery-agenda` (a Google Doc, so the
222
+ technician can read it from any device mid-call). The command updates the Doc
223
+ that is already there **in place** — the same link — and never creates a
224
+ second one. One agenda per call, everywhere. The Drive
196
225
  copy is for running the call; the build-repo file remains the versioned
197
226
  source. Leave the `Type:` line out of the Drive copy — it is a repo
198
227
  convention.