Skip to content

JasonColapietro/suede-creator-skills / skills/suede-newsroom

suede-newsroom

Verified

Suede Newsroom Skills. editorial pipeline with six role contracts, an inspectable campaign record, distinct distribution arguments, and performance feedback. Use when asked "build an AI content team" or to repair duplicated, generic, or unsourced agent content. NOT FOR: topics (suede-content-strategy), channel calendars (suede-social), scheduling (suede-marketing-loops), code lanes (suede-agent-teams), or flagship writing (suede-ship-copy).

FileKindSize
CARD.mdmarkdown4.5 KB
SKILL.mdskill md16.1 KB
agents/openai.yamlconfig456 B
evals/evals.jsonconfig7.9 KB
references/campaign-record.mdmarkdown4.9 KB
references/entryways.mdmarkdown5.6 KB
references/founder-content-loop.mdmarkdown18.7 KB
references/role-contracts.mdmarkdown7.9 KB
references/worked-contract.mdmarkdown2.2 KB

references/founder-content-loop.md

View on GitHub
# Founder-led content loop

Use this mode when a founder interview, voice note, or operator debrief is the
recurring source, or when personal and company accounts need different editorial
jobs. It extends the newsroom pipeline; it does not replace the six role
contracts, campaign record, evidence gates, entryways, approval boundary, or
Keep/Test/Stop review in `SKILL.md`.

## Contents

1. [Intended use](#intended-use)
2. [How it works](#how-it-works)
3. [Mode contract](#mode-contract)
4. [Founder intake](#founder-intake)
5. [Weekly operating rhythm](#weekly-operating-rhythm)
6. [Account lanes](#account-lanes)
7. [Distribution and collision rules](#distribution-and-collision-rules)
8. [Measurement](#measurement)
9. [Worked examples](#worked-examples)
10. [Routing boundaries](#routing-boundaries)
11. [Pilot size](#pilot-size)
12. [Output contract](#output-contract)

## Intended use

Use founder-led mode to install a repeatable editorial loop when the most useful
raw material lives in a founder's weekly judgment rather than in a backlog of
finished drafts. Use it when one or more of these apply:

- a founder can supply a recurring interview, voice note, or operator debrief;
- personal and company accounts need distinct editorial jobs;
- the team needs a written system that survives restarts and handoffs;
- claims, approvals, and performance lessons must stay inspectable; and
- the desired result is recognition, qualified inbound, product adoption, or
  community depth rather than content volume by itself.

Use the standard newsroom without this mode when neither a recurring founder
source nor an account-lane distinction applies. Route topic pillars to
`suede-content-strategy`, platform listening and calendars to `suede-social`,
stateful or retrying scheduled stages to `suede-marketing-loops`, analytics
instrumentation to `suede-analytics`, and platform publishing or scheduling to
`suede-social` or the applicable platform-specific skill.

This mode prepares an approval-ready editorial package. It does not choose a
public identity, publish a post, invent proof, or turn one successful campaign
into a permanent rule.

## How it works

The loop converts one recurring source into a sequence of written decisions:

```text
setup -> founder intake -> signal -> research -> angle -> flagship
      -> distinct account assets -> human approval -> measurement -> playbook
```

1. **Setup fixes the operating constraints.** The primary outcome, account
   lanes, source format, approval owner, measurement window, and review capacity
   live in `.agents/newsroom/founder-loop.md`.
2. **Intake extracts judgment without treating it as proof.** Firsthand
   observations, decisions, voice markers, proof leads, and claims to verify are
   separated in the campaign record.
3. **The newsroom narrows the material.** The scout selects at most three
   candidates, research verifies consequential claims, and strategy chooses one
   angle.
4. **Distribution gives every asset a different job.** The founder lane earns
   recognition through a mechanism, workflow, risk, critique, or lesson. The
   company lane makes the work inspectable through proof, a scoped result, a
   documented workflow, or a grounded model.
5. **The editor checks the set together.** Every asset must stand alone, use one
   supported entryway, and pass claim, opening, and entryway collision checks.
6. **A human releases the work.** The package names the exact copy, source,
   media, visible identity, and approval decision. The newsroom does not publish
   or schedule it.
7. **Measured results update the playbook.** One promising result becomes a
   Test. A Keep requires support from at least two campaigns and human approval.

The durable setup file carries the rules for the whole loop. Each campaign file
carries one idea through the stages. Source recordings, transcripts, drafts,
and media live beside those records and are linked by path.

## Mode contract

The founder supplies judgment, firsthand observations, decisions, language, and
proof leads. The newsroom turns that material into an evidence-backed campaign.
The founder is a source of experience, not automatic verification for a factual
claim.

Complete the mode setup when these fields are written:

```text
setup path:           .agents/newsroom/founder-loop.md
primary outcome:       the business or reputation action the loop should create
founder voice:         who speaks and which account represents that person
company proof lane:    which account represents the company or product
durable themes:        supplied upstream; 3-5 only when strategy supports that range
weekly source:         interview | voice note | operator debrief
capture window:        duration and owner
approval owner:        person who approves angle, claims, copy, account, and media
measurement window:    when the performance record becomes reviewable
capacity:              maximum campaigns and public assets the human can review
```

Write those fields to `.agents/newsroom/founder-loop.md`. Each campaign record
points to that file through `mode-setup-path`, so a restart never asks the next
role to reconstruct an identity, owner, capacity, or outcome from conversation.
Update the setup file only after the human approves the changed field.

Carry supplied values exactly. Write `open` for a missing duration, identity,
measurement window, or capacity. A pilot may recommend a starting test only
after the open value is shown; the recommendation stays labelled `proposed` and
does not become the operating rule without human approval.

An empty primary outcome, approval owner, or account lane halts setup. Ask for
the missing field; do not choose a public identity or success measure by
inference.

## Founder intake

Create the `Founder intake` section in the campaign record before the signal
stage. Store the source beside the record and write its path into `source-path`.
The record carries extracted decisions; it does not carry a full transcript.

Capture these prompts in the founder's natural words:

1. What happened, changed, broke, or became clearer this week?
2. Which decision did you make, and what tradeoff drove it?
3. What have you observed firsthand?
4. What is the industry getting wrong or leaving unexplained?
5. What could a stranger inspect: source, screenshot, demo, dataset, workflow,
   result, or before-and-after?
6. Which statement still needs verification?
7. Which phrase, analogy, or opinion sounds unmistakably like you?
8. Which detail is private, off-limits, or too early to publish?

Sort the answers into five buckets:

| Bucket | Meaning | Next owner |
|---|---|---|
| Firsthand observation | what the founder directly saw or did | scout and strategist |
| Decision or lesson | the founder's judgment and reasoning | strategist and writer |
| Voice marker | phrasing, analogy, rhythm, or conviction to preserve | writer and editor |
| Proof lead | an object or source that may support a claim | researcher |
| Claim to verify | a factual statement with consequence | researcher |

`Proof lead` and `claim to verify` remain separate until the researcher supplies
the source, scope, date, and limits. A founder-supplied result, customer outcome,
ranking, percentage, or market claim is not verified because it appears in a
recording.

The intake is complete when every answer appears in one bucket, `off-limits` is
explicit, and the source path resolves to the stored recording or transcript.

## Weekly operating rhythm

This rhythm is a manual editorial playbook. Days are labels the operator may
shift; they are not a scheduler or an instruction to publish.

### 1. Select inputs

Read the durable themes, current conversation evidence, last performance review,
and Stop list. Generate a bounded prompt set for the founder interview. The
prompt set may contain 5-10 questions when the capture window can answer them;
cut it until every question fits the stated window.

The signal scout still returns at most three campaign candidates. A long topic
list is interview preparation, not approval to open ten campaigns.

Complete this stage when every selected question maps to a durable theme, a
current conversation, a customer question, or a prior Test hypothesis.

### 2. Capture founder judgment

Run the interview or ingest the voice note. Fill the Founder intake section and
mark claims that need research. Preserve exact wording only when the recording
supports it; attach a timestamp to any direct quote.

Complete this stage when all intake fields pass the criteria above.

### 3. Run the newsroom

Move the record through signal, research, angle, flagship, distribution, and
review using the existing role contracts. One founder intake can yield several
signal candidates, but each approved campaign gets its own record and exactly
one recommended angle.

Complete this stage when the editor returns an approval-ready package or the
record is halted at the field that blocks it.

### 4. Release through the human boundary

Present final copy, supporting source, media, target account, visible identity,
and the exact decision required. The newsroom publishes and schedules nothing.

Complete this stage when the human has approved or rejected every public asset;
an unanswered item remains pending.

### 5. Learn

After the measurement window closes, update the campaign performance section
and run Keep/Test/Stop. One result remains a Test. The human approves any
standing-rule change.

Complete this stage when every proposal names its supporting campaigns and the
approval state is recorded.

## Account lanes

The accounts share evidence and positioning. They do not share a central claim,
opening, or shortened version of the same draft.

### Founder recognition lane

The founder account makes the operator's judgment recognizable. Its asset should
teach one of these:

- a mechanism the founder can explain from experience;
- a workflow or decision rule another operator can use;
- a risk and the condition that triggers it;
- a critique supported by evidence;
- a lesson from a named decision or failure.

The asset is complete when the reader can name what the founder believes or does
differently and the packet supports every consequential claim.

### Company proof lane

The company account makes the work inspectable. Its asset should carry one of
these:

- a proof object a stranger can check;
- a result with scope, conditions, and measurement source;
- a documented workflow or product behavior;
- a compressed diagram or model grounded in the campaign;
- a company update whose relevance does not depend on the founder post.

The asset is complete when it stands alone, names what can be inspected, and
does not ask the reader to trust the founder's authority as proof.

Use only the lanes the user supplied. A solo founder without a company account
can run the recognition lane alone. A company-led campaign without a public
founder can run the proof lane alone.

## Distribution and collision rules

Apply the normal entryway and standalone scoring rules to every asset. Add
`account lane` to each row in the Distribution section:

```text
surface | visible identity | account lane | entryway | central claim |
standalone reason | score | claims used | approval state
```

Reduce the full cross-account set to one sentence per asset, then run claim,
opening, and entryway collision checks. The account name does not make two
colliding assets distinct.

Resolve a collision by assigning an unused supported entryway, cutting the
weaker asset, or recording a deliberate repeat with a reason. Rewording the same
claim does not resolve it.

The set is complete when every asset scores 3, every lane is named, and the
collision field is clean or records the human-approved resolution.

## Measurement

Set the primary outcome before choosing metrics. Record the goal-linked action
first and platform diagnostics second.

| Outcome | Goal-linked actions to consider |
|---|---|
| Qualified inbound | relevant replies, introductions, calls, demos, or inquiries |
| Category recognition | invitations, citations, mentions, or follows from the named audience |
| Product adoption | artifact views, qualified visits, trials, installs, or activated accounts |
| Community depth | substantive replies, repeat contributors, or useful user questions |

Use only measures the platform or connected analytics actually expose. Reach,
likes, and follower count can diagnose distribution; they do not replace the
primary outcome.

Compare by campaign, account lane, entryway, and surface. A Keep needs two or
more campaigns and continued strategy fit. One promising result is a Test. A
Stop requires the thresholds in `SKILL.md` Step 7.

## Worked examples

These examples are illustrative. They show the decisions and artifacts the mode
returns; they are not claims about a real campaign.

### Example 1: Weekly voice note for qualified inbound

**Request:** "The founder and company accounts are already approved in the mode
setup. I approve every asset, can review two public assets per week, and review
results after seven days. The goal is qualified product inquiries. This week's
voice note says we saw teams diagnose onboarding symptoms without a decision
rule, and I attached the checklist we use."

**What the mode does:**

1. Writes `qualified product inquiries` as the primary outcome and records the
   approved accounts, approval owner, seven-day measurement window, and two-
   asset review capacity in the setup file.
2. Classifies the founder's diagnosis as firsthand observation and judgment.
   The checklist is a proof lead until the researcher verifies that the shared
   artifact matches the description.
3. Selects one angle: onboarding fails when teams diagnose symptoms without a
   decision rule.
4. Gives the founder asset a **mechanism** entryway: explain the decision rule
   and why it changes the diagnosis.
5. Gives the company asset a **proof** entryway: show the inspectable checklist,
   its scope, and what it does not establish.
6. Rejects a second post claiming the checklist improved conversion unless a
   measurement source, comparison period, and conditions are available.
7. Sends both assets to one approval queue, then measures qualified replies,
   introductions, demos, or inquiries before using reach as a diagnostic.

**Result:** two assets share one evidence packet but make different central
claims. The founder is recognizable for a decision method; the company makes
the method inspectable.

### Example 2: A time-sensitive industry conversation

**Request:** "Use the existing approved mode setup. A new industry policy is
being debated this week. I attached the policy source, my operator debrief on
the operational risk, and the company's documented workflow. Build the package
while the conversation is current."

**What the mode does:**

1. The scout records the conversation source and decay window. If the window
   expires before research and approval finish, the timely angle is rejected or
   reframed as durable analysis.
2. The researcher separates what the policy text proves from the founder's
   interpretation and verifies the company workflow independently.
3. The founder asset takes the **risk** entryway: name the failure condition and
   the operator decision it changes.
4. The company asset takes the **workflow** entryway: show the ordered process
   and the evidence for each consequential claim.
5. The collision gate compares both central claims and openings. Rephrasing the
   founder warning as company copy does not count as a distinct asset.
6. The human approves the final wording, visible identity, sources, and media
   before any publishing skill receives the package.

**Result:** the system participates while the topic is live without converting
speed into unsupported claims or duplicated account posts.

### Example 3: Solo founder with one public account

**Request:** "I have one approved founder account and no company account. The
goal is category recognition, I approve every asset, I can review one public
asset per week, and performance becomes reviewable after seven days. Start from
the attached operator debrief."

**What the mode does:**

1. Activates only the supplied founder recognition lane; one valid lane
   satisfies the mode contract, and the absent company lane stays unavailable.
2. Records category recognition, the founder as approval owner, the one-asset
   weekly review capacity, and the seven-day measurement window.
3. Runs the debrief through intake, research, one selected angle, one supported
   entryway, standalone scoring, and human approval.
4. Records company-grade proof objects as possible evidence, not as instructions
   to create another public account.
5. Logs the campaign as measured or explicitly `unmeasured`; it does not promote
   a lesson into Keep until the threshold is met.

**Result:** the smallest valid installation is one accountable lane and one
inspectable record. The system expands only after the missing operating fields
are supplied and approved.

If any example omitted its primary outcome, approval owner, usable account lane,
or actual source, setup would halt at that field. A hypothetical description of
a future voice note or debrief is not campaign evidence.

## Routing boundaries

- Pillar discovery, topic ownership, and portfolio allocation belong to
  `suede-content-strategy`. This mode consumes the approved themes.
- Listening, platform mix, calendar design, and surface conventions belong to
  `suede-social`. This mode consumes those inputs and assigns distinct arguments.
- Scheduled execution, state, retries, idempotency, and stop conditions belong
  to `suede-marketing-loops`.
- Analytics instrumentation and attribution belong to `suede-analytics`.
- Publishing, scheduling, replies, and other account actions on any platform,
  Instagram included, route to `suede-social`. Each still requires approval of
  exact content and visible identity, plus an authorized platform tool.

Route the missing capability; keep the founder intake, account lanes, evidence
record, and editorial gates in this mode.

## Pilot size

Set the pilot count from the capacity field. When capacity is `open`, return a
`proposed` pilot shape without an exact campaign or asset count and wait for the
human to approve the operating limit. The generic newsroom's three-idea
calibration is not a founder-mode capacity default.

The pilot is complete when its approved campaign count has crossed every gate
and each campaign is measured or explicitly `unmeasured`.

## Output contract

Return the installed operating system in this shape:

```text
Mode setup:
Founder intake contract:
Weekly operating rhythm:
Newsroom role map:
Founder recognition lane:
Company proof lane:
Campaign record changes:
Human approval queue:
Measurement and Keep/Test/Stop:
Routed dependencies:
Capacity-sized pilot:
Open fields:
```

The output is complete when every supplied account has one job, every weekly
stage ends on a checkable criterion, the campaign record can carry the founder
intake, and publishing remains outside the newsroom.
writing-skillsobra/superpowersUse when creating new skills, editing existing skills, or verifying skills work before deploymentStars: 323kUpdated Claude CodeCodexVerifiedInstall
xlsxanthropics/skillsUse this skill any time a spreadsheet file is the primary input or output. This means any task where the user wants to: open, read, edit, or fix an existing .xlsx, .xlsm, .xltx, .csv, or .tsv file (e.g., adding columns, computing formulas, formatting, charting, cleaning messy data); create a new spreadsheet from scratch or from other data sources; or convert between tabular file formats. Trigger especially when the user references a spreadsheet file by name or path — even casually (like "the xlsx in my downloads") — and wants something done to it or produced from it. Also trigger for cleaning or restructuring messy tabular data files (malformed rows, misplaced headers, junk data) into proper spreadsheets. The deliverable must be a spreadsheet file. Do NOT trigger when the primary deliverable is a Word document, HTML report, standalone Python script, database pipeline, or Google Sheets API integration, even if tabular data is involved.Stars: 198kUpdated Claude CodeCodexVerifiedInstall
shipping-and-launchaddyosmani/agent-skillsPrepares production launches. Use when preparing to deploy to production, or when asking what needs to be in place before shipping. Use when you need a pre-launch checklist, when setting up monitoring, when planning a staged rollout, or when you need a rollback strategy.Stars: 119kUpdated Claude CodeCodexVerifiedInstall
childhoodthedotmack/claude-memBabysit a shipped release after merge, publication, or deployment. Verify published packages, production behavior, installed plugins, and late CI or user reports until the observation window is clean; use babysit for PR readiness before release.Stars: 118kUpdated Claude CodeCodexVerifiedInstall