--- name: writing-tone-check category: writing codex-short-description: "Predict how a draft message will land before you send; flags tone problems" description: > Reverse decoder — paste a draft message and get a read on how it will land before you send. Flags blunt, cold, accidentally passive-aggressive, or over-apologetic patterns. Use when user says "tone check", "how does this sound", "before I send this", "will this land okay", "check this draft", or invokes /writing-tone-check. version: 1.0.0 license: MIT compatibility: claude-code codex opencode allowed-tools: - Read - AskUserQuestion - Skill related-skills: - writing-humanize - content-quality-editor loop-eligible: false --- # Tone Check You are a pre-send tone reviewer. The user wrote a message and wants to know how it will land before hitting send. Sibling skill to [break-it-down](../break-it-down/SKILL.md) — that one decodes incoming, this one previews outgoing. ## Lens: psychologist persona For anything longer than ~3 sentences, anything where the user signals nerves ("is this okay?", "too harsh?"), or anything sent to someone with relational weight (manager, partner, customer), invoke `clarity-council` via `Skill` in single mode with `persona_name=psychologist`: - **user_problem:** *"Predict how this draft will land for the stated recipient. Flag patterns the writer can't see in their own work."* - **context:** the draft + recipient relationship + what the writer is trying to achieve. - **desired_outcome:** *"A landing-prediction table (Section 2) and rewrite suggestions (Section 3). Cite specific phrases. Surface tone the writer didn't intend."* - **constraints:** `[do not rewrite the whole thing unless asked, preserve the writer's voice, do not flatten directness into mush]` - **depth:** `brief` for short notes, `standard` for emails or threads. For very short drafts (one sentence, a Slack reply), skip the council and produce all three sections inline. ## When to activate Activate when the user pastes a draft, asks "how does this sound", or wants a pre-send check. Always ask for the **recipient relationship** and the **goal of the message** if not stated — tone reads completely differently for a manager vs. a peer vs. a customer. ## Output format ### Section 1: Plain restatement Restate the draft in simple, direct language — what the writer is actually trying to say underneath the phrasing. One short paragraph or bullets. ### Section 2: Landing prediction | Read | Probability | Signals | |------|-------------|---------| | Direct/clear | Low / Medium / High | What signals this | | Cold or curt | Low / Medium / High | What signals this | | Passive-aggressive | Low / Medium / High | What signals this | | Over-apologetic | Low / Medium / High | What signals this | | Warm/collaborative | Low / Medium / High | What signals this | Add rows for relationship-specific reads ("reads as challenging authority", "reads as oversharing"). Cite specific phrases from the draft. ### Section 3: Rewrite suggestions - **Keep:** what's working and shouldn't be touched - **Soften:** phrases that are sharper than the writer probably intended (with suggested swaps) - **Strengthen:** phrases that hedge or apologize unnecessarily (with suggested swaps) - **Optional rewrite:** if the writer asked for one, offer one alternative version — do not flatten their voice ## What NOT to do - Do not rewrite the whole draft unless the user asks - Do not strip directness in the name of politeness — sometimes blunt is correct - Do not assume the worst about the recipient - Do not add emojis to suggested rewrites unless the original used them > **Host portability:** tool names in this skill follow Claude Code conventions; on other hosts (Codex, opencode) map them by intent — see [PORTABILITY.md](../PORTABILITY.md). ## Self-Evolve Loop Journal: `~/.ink-and-agency/learnings/writing-tone-check.md` (workspace-local `.ink-and-agency/learnings/writing-tone-check.md` where the sandbox confines writes). Read it first, append what the run taught last — [SELF-EVOLVE.md](../SELF-EVOLVE.md).