comm — Communication Consultant Skill

A Claude Code skill that acts as a communication consultant. Maintains a registry of per-person profiles, ships with communication theory references (DISC, attachment, conflict modes, situational leadership, etc.), and returns strategy + register guidance before you draft any message to a specific person.

Install

1. Clone: git clone <repo> ~/comm-skill
2. Symlink: ln -s ~/comm-skill ~/.claude/skills/comm
3. (Optional) Copy comm-config.example.json to ~/.claude/skills/comm/comm-config.json and set data_root to a writable directory holding your profiles. Default is <skill>/data/. The config also accepts pr_review_policy ("binary" | "non-blocking-allowed") — see SKILL.md for semantics.

Getting Started

After installing, run /comm onboard — it walks you through the setup that matters:

1. Builds your self-profile (your communication defaults, strengths, watch-outs). This is what lets the skill override naive "just mirror their style" advice — the highest-value step.
2. Creates your first one or two recipient profiles (e.g. your manager, a close peer).
3. Optionally profiles the cultures on your team so cross-cultural calibration works.

The skill ships empty by design: no profiles of real people, only frameworks and templates. Everything about you and your colleagues is created locally during onboarding and stays in your data root — keep it private; profiles hold candid notes.

Usage

• /comm onboard — first-run setup (build your self-profile + first profiles)
• /comm — freeform consult ("help me reply to X about Y")
• /comm calibrate <person> [situation] — strategy + register guidance
• /comm profile list|show|new|edit|refresh [person|all]

See SKILL.md for the full surface.

Make it fire automatically (recommended)

In Claude Code the skill already auto-triggers from its description when you ask to draft, reply to, or calibrate a message to a specific person. To make that reliable — so the agent calibrates before drafting instead of defaulting to mirroring the recipient's recent messages — add a rule to your CLAUDE.md (or AGENTS.md / GEMINI.md):

Recipient-calibrated communication. Any message addressed to a specific person (Slack, email, LinkedIn, PR or Linear comment, doc reply): invoke the /comm skill BEFORE drafting. Style-mirroring from recent messages is never the default — calibrate first.

That one line is what turns the skill from "available when asked" into "always consulted before you send something to a real person." Adjust the channels and emphasis to taste.