| name | nemawashi-reply |
| description | Use when user needs help replying to someone, crafting a message, or deciding what to say - references profiles to tailor communication strategy |
Reply Strategist
Help the user craft optimal replies and messages by leveraging profile data about the recipient.
When to Use
- User says "how should I reply to X?"
- User says "what should I say to X about Y?"
- User asks for help writing a Slack message, email, or any communication
- User is dealing with a difficult conversation
Process
Step 1: Identify Context
Determine:
- Who is the recipient?
- What is the topic/situation?
- Where will this be sent? (Slack, email, in-person, etc.)
- What outcome does the user want?
Step 2: Load Profile (selective)
Read the recipient's profile from PROFILE_DIR/<person-name>/. The profile is split into a slim index (profile.md) plus per-framework files (frameworks/<slug>.md). Load only what the current situation needs:
Always load:
profile.md — Basic Info, Core Pattern, Framework Summary (the synthesis).relationship.md — User's relationship and approach strategy.contradictions.md — Known contradictions (relevant for difficult conversations; cheap to load).
Conditionally load (per Step 3's situation mapping). The framework slugs referenced below are the canonical set registered in skills/nemawashi-analyze/FRAMEWORKS.md; add a row + a mapping entry below whenever a new framework lands there:
| Situation | Always load | Add if signals suggest |
|---|
| When Requesting | frameworks/core-motivators.md, frameworks/cognitive-biases.md | frameworks/transactional-analysis-ta.md (status-charged ask) |
| During Conflict | frameworks/thomas-kilmann-tki.md, frameworks/defense-mechanisms.md | frameworks/transactional-analysis-ta.md (CP↔AC dynamic), frameworks/attachment-style.md (trust rupture) |
| When Reporting | frameworks/defense-mechanisms.md, frameworks/core-motivators.md | frameworks/cognitive-biases.md (loss-aversive recipient) |
| Routine Collaboration | frameworks/transactional-analysis-ta.md, frameworks/core-motivators.md | frameworks/attachment-style.md (delegation/trust question) |
If a frameworks/<slug>.md does not exist or its frontmatter says confidence: Data Gap, skip it.
Recent facts. Read only the most-recent fact entries (last ~10) for context continuity — the framework files already encode the analysis. Read facts.jsonl (newer) / facts.md (legacy); if both exist, read both and merge.
If no profile exists: Tell the user and offer two options:
- Run
nemawashi-collect first for a data-informed reply.
- Proceed without a profile (generic advice only).
If frameworks/ is missing but profile.md exists (pre-split profile): tell the user to run nemawashi-analyze (or /supernemawashi:nemawashi-migrate to bulk-upgrade), then proceed loading profile.md only.
If profile is stale (last_updated > 30 days): suggest updating, but proceed with existing data if the user wants a quick reply.
Step 3: Analyze the Situation
Map to situation category. Determine which of the 4 categories defined in using-supernemawashi → Situation Categories applies.
Read the DO/DON'T rules for that situation category from each loaded frameworks/<slug>.md file's ## Rules → ### <situation> block. These rules are backed by psychological framework analysis (defense mechanisms, conflict modes, ego states, motivators, cognitive biases) — use the per-rule [signal: ...] tags to explain your reasoning to the user. The cross-framework synthesis is in profile.md → Core Pattern.
Based on profile data, also assess:
- Recipient's likely reaction — Given their behavioral patterns and framework classifications, how will they probably respond to this topic?
- Timing — Is now a good time based on their patterns? (e.g., avoid mornings if they're irritable early)
- Channel fit — Is the chosen channel optimal for this person and topic?
- Political context — Should anyone else be consulted or CC'd first?
Step 4: Propose Strategy
Present the user with:
Recommended approach:
- Tone (direct/indirect, formal/casual)
- Structure (lead with data? ask a question? frame as their idea?)
- Key phrases to use or avoid
- Timing recommendation
Risks and mitigations:
- What could go wrong
- How to handle likely pushback
- Fallback positions
Step 5: Draft Message
Write 2-3 draft variations:
- Safe option — Minimal risk, conventional approach
- Strategic option — Optimized for the desired outcome based on profile analysis
- Direct option — Straightforward, when the situation calls for it
For each draft, annotate WHY specific choices were made based on profile data. Example:
"Leading with the data point because profile indicates this person responds better to numbers than narratives."
Step 6: User Selects and Refines
Let the user pick a draft or mix elements. Refine until they're satisfied.
Do NOT send the message. The user decides when and whether to send. If asked to send, use the appropriate MCP tool (Slack, Gmail, etc.) only with explicit user confirmation.
Handling Difficult Situations
For conflict, criticism, or high-stakes conversations:
- Reference contradictions.md if the person has a history of inconsistent statements
- Suggest de-escalation language when tension is detected
- Recommend involving allies (from relationship.md approach strategy)
- Propose waiting/cooling-off if the situation is heated
Key Principles
- Profile-informed, not profile-dependent — Useful even without a profile, exceptional with one.
- User has final say — Never send without explicit confirmation. Proposals only.
- Explain the reasoning — Every suggestion should reference specific profile data so the user understands WHY.
- Multiple options — Always provide choices. Don't prescribe a single "right" answer.
- Context-aware — Consider the full situation: timing, channel, audience, political dynamics.
