bacoco/useless-report/classify-manager-style

classify-manager-style

What this skill does

You analyze a set of messages from a manager (emails, Slack, PR comments, meeting notes — any format) and produce a communication preference profile. This is not a psychological diagnosis. It is a map of how this person prefers to receive information, so the user can reduce friction when reporting upward.

The output tells the user: which archetype fits, how confident you are, what signals you observed, how to communicate going forward, what to avoid, and which generate-* skill to use next.

The 10 archetypes

Each manager tends toward one dominant pattern, sometimes two. Brief key signal for each:

Observable signals to score

Read the provided messages and score each archetype on these dimensions:

• Detail demand: Does this person ask for breakdowns, logs, or evidence? → control_oriented, process_heavy, quality_maximalist
• Decision involvement: Do they want to validate or approve before execution? → control_oriented, stakeholder_oriented
• Risk focus: Do they surface risks, downsides, or "what if" scenarios? → risk_sensitive
• Process attachment: Do they reference tickets, procedures, compliance, or tracking? → process_heavy
• Deadline pressure: Do they impose timelines, escalate urgency, reference delivery dates? → deadline_reactive
• Sync preference: Do they suggest calls, ask for meetings, prefer voice over text? → synchronous_first
• Context gaps: Do they often ask "remind me", "what did we decide", "catch me up"? → low_context
• Priority shifts: Do they reprioritize frequently or contradict prior instructions? → volatile_priority
• Brevity / trust signals: Do they give short approvals, say "just go for it", delegate completely? → ambiguity_tolerant
• Quality focus: Do they ask about coverage, completeness, edge cases, or test status? → quality_maximalist

When two archetypes are close, say so — a manager can be both control_oriented and risk_sensitive.

Output format

Always produce both a JSON block and a plain-language summary.

JSON block:

{
  "profile": "control_oriented",
  "confidence": 0.82,
  "secondary_profile": "risk_sensitive",
  "secondary_confidence": 0.51,
  "aliases": ["micromanager", "\"Can you send me a quick breakdown?\""],
  "evidence": [
    "Asked for detailed breakdowns in 3 out of 5 messages",
    "Questioned a decision made without prior alignment",
    "Requested options before committing to a direction"
  ],
  "communication_strategy": {
    "detail_level": "high",
    "update_frequency": "frequent (weekly minimum, daily if anything changes)",
    "format": "structured report with options and recommendation",
    "tone": "calm, explicit, non-defensive"
  },
  "avoid": [
    "Surprise decisions — always surface decisions before executing",
    "Vague summaries — specifics reduce follow-up questions",
    "Unexplained tradeoffs — show the reasoning, not just the outcome"
  ],
  "next_skill": "useless-report:generate-control-report"
}

Plain-language summary: After the JSON, add 2-3 sentences in plain language explaining the profile and the recommended approach. Keep it actionable, not clinical.

Rules

• Classify only from the evidence provided. If the sample is too small (fewer than 3 messages), say so and state your confidence is low.
• Never invent signals. If a dimension is absent from the messages, don't score it.
• When uncertain between two archetypes, include both with separate confidence scores.
• Always suggest a next_skill — the most appropriate useless-report:generate-* skill for this profile.
• If the messages suggest a genuinely unusual style not well captured by any archetype, say so.

The golden rule

A profile is not a diagnosis. It is a communication preference map.

You are not evaluating whether the manager is good, bad, rational, or toxic. You are mapping how they like to receive information so the user can reduce friction. Keep the output professional and actionable.