signal

name	signal
description	Track signals from the enterprise agentic AI landscape — analyst reports, production case studies, failure postmortems, vendor announcements — and surface contradictions or confirmations with the Mentilead experiment. Three modes: log a signal (under 60 seconds), review last 60 days grouped by classification, or generate content angles bridging enterprise signals to personal experiment findings. Feeds the practitioner's dual perspective.
argument-hint	What signal did you see?

GrowthOS Signal Monitor

Step 0: Detect & Route

Before starting any mode, determine the current state and recommend the next logical step.

Read Current State

1. Read marketing/MEMORY.md and marketing/STATUS.md
2. If marketing/ does NOT exist: tell the user to run the init skill first and stop
3. Read references/experiment-thesis.md — required by the Experiment Layer rules in CLAUDE.md
4. Check if marketing/experiment/signals.md exists. If not, offer to create it from templates/signals.md
5. Read experiment_chapter from STATUS.md frontmatter (default: "Chapter 1: Foundation")
6. Count signals by classification (contradiction, confirmation, unknown-yet) and count unused signals (where Used in is —)

Slash Command Fast-Path

If invoked via slash command, skip the options menu and go directly to Step 1 (Log Signal).

Present Options

If the user's intent is not clear from their message, present the dashboard:

Signal Monitor — {experiment_chapter}:

Signals logged: {total}
Contradictions: {N} | Confirmations: {N} | Unknown: {N}
Unused in drafts: {unused}/{total}

1. Log Signal — capture an enterprise AI signal (under 60 seconds)
2. Signal Review — last 60 days grouped by classification
3. Content Angles — generate angles from unprocessed signals

Route to the appropriate step based on the user's choice or detected intent.

Step 1: Log Signal

Designed for speed — the user should be able to log a signal in under 60 seconds.

Gather Information

Ask only what's needed (speed-oriented):

• "What signal did you see?" (1-3 sentences)
• "Where from?" (source — report name, conference, publication, conversation pattern)
• If classification is not obvious from context: "Does this confirm or contradict anything from your experiment? Or not sure yet?"

Confidentiality Guard

Never log client names or confidential details from enterprise conversations. If the user's input appears to contain a specific client name for enterprise-conversation type signals, ask them to generalize before logging. Capture the pattern or insight only.

Auto-Classify

From the user's answers, classify:

• type: one of production-deployment, failure-cancellation, vendor-announcement, analyst-report, enterprise-conversation, regulatory
• classification: one of confirmation, contradiction, unknown-yet
• content_value:
  • contradiction → always high
  • confirmation backed by specific data → medium
  • All other cases → low

Determine Signal ID

Count existing ### SIG- entries in marketing/experiment/signals.md and increment by 1. Format as SIG-{NNN} with zero-padded three digits (e.g., SIG-001, SIG-012).

Append Entry

Append to marketing/experiment/signals.md:

### SIG-{NNN} — {YYYY-MM-DD}
- **Source:** {where — report name, conference, publication, conversation pattern}
- **Type:** {type}
- **Classification:** {classification}
- **Chapter:** {experiment_chapter from STATUS.md}
- **Signal:** {1-3 sentences — the pattern or insight, never client names or confidential details}
- **Experiment connection:** {how this relates to the thesis, or "—"}
- **Content value:** {high / medium / low}
- **Used in:** —

Handle Contradictions

If classification is contradiction:

• Set content_value: high regardless of other factors
• Prompt: "This contradicts the experiment's findings. Quick content idea? (one sentence, or skip)"
• If provided, append - **Content idea:** {idea} to the entry

Confirm to User

Show the logged entry back to the user:

Logged SIG-{NNN}:
- Source: {source}
- Classification: {classification}
- Content value: {content_value}
{if contradiction: "Contradiction flagged — high content value."}

Session Log & State Update

1. Append to marketing/logs/{YYYY-MM-DD}.md:

   - **Signal Monitor** — Logged SIG-{NNN} ({classification}): {one-line summary}
   

2. Update marketing/STATUS.md: set last_updated to today's date, preserve experiment_chapter

Step 2: Signal Review (Read-Only)

Review signals from the last 60 days, grouped by classification.

Gather Data

1. Read marketing/experiment/signals.md
2. Filter entries for the last 60 days (parse date from ### SIG-{NNN} — {YYYY-MM-DD} headings)

Present Review

Group by classification — contradictions first, then confirmations, then unknown-yet. Mark [unused] for entries where Used in is —.

Signal Review — Last 60 Days:

Contradictions:
- SIG-{NNN} ({date}) [{type}] — {signal summary} {[unused]}

Confirmations:
- SIG-{NNN} ({date}) [{type}] — {signal summary} {[unused]}

Unknown:
- SIG-{NNN} ({date}) [{type}] — {signal summary} {[unused]}

Prompt Reclassification

If any unknown-yet signals are older than 30 days, prompt the user:

These signals have been unclassified for over 30 days — ready to classify?
- SIG-{NNN} ({date}): {signal summary}

If the user provides a classification, update the entry's Classification and Content value fields accordingly.

Session Log

Append to marketing/logs/{YYYY-MM-DD}.md:

- **Signal Monitor** — Signal review: {total} signals in last 60 days ({contradictions} contradictions, {confirmations} confirmations, {unknown} unknown)

No STATUS.md write — this is read-only mode.

Step 3: Content Angle Generator (Read-Only)

Generate content angles from unprocessed signals, bridging enterprise signals to experiment findings.

Gather Data

1. Read references/experiment-thesis.md — required for framing angles against the thesis
2. Read marketing/experiment/signals.md — filter for unprocessed signals (Used in is —)
3. Read marketing/experiment/observations.md — for matching experiment findings

Select Signals

Pick up to 3 signals for angle generation, prioritized:

1. Contradictions (highest content value)
2. Signals with a Content idea field
3. Confirmations backed by specific data

Generate Angles

For each selected signal, present:

Content Angle from SIG-{NNN}:

Enterprise says: "{signal summary}"
My experiment shows: "{relevant observation or finding}"
Angle: "{proposed title or framing}"
Platform: {Substack for detailed analysis, LinkedIn for punchy contradiction}
Related observations: OBS-{NNN}, ...

Close

After presenting angles:

Want to draft any of these? Run the publisher skill with the signal as context.

Session Log

Append to marketing/logs/{YYYY-MM-DD}.md:

- **Signal Monitor** — Generated {N} content angles from unprocessed signals

No STATUS.md write — this is read-only mode.

Reference Table

Mode	References Read	User Files Read	User Files Written
Step 0	references/experiment-thesis.md	MEMORY.md, STATUS.md, signals.md	—
Step 1	—	STATUS.md	signals.md, STATUS.md, logs/{today}.md
Step 2	—	signals.md	logs/{today}.md
Step 3	references/experiment-thesis.md	signals.md, observations.md	logs/{today}.md