--- description: Salary Negotiation — triggered at offer time. Pre-built scripts for common scenarios (geo-discount pushback, RSU acceleration, competing-offer leverage, signing-bonus ask, base-vs-equity tradeoffs). Tier 4 main session. Voice-strict. --- # personal-salary-negotiation Negotiation drafting is irreversible-broadcast-adjacent and high-stakes future-employer-content. **Tier 4 main, no exceptions, no delegation.** ## Trigger phrases - "Help me negotiate this offer" - "[Company] offered $X — how do I push back?" - "Salary negotiation for [role]" - `/negotiate` ## Inputs - Offer details (Josh provides — base, bonus structure, equity, sign-on, benefits, location adjustments) - `~/.career/applications/{slug}/` — full application context - `~/.career/companies/{slug}/brief.md` — company comp signal - `~/.career/voice-profile.md` - Levels.fyi data for the role + level + company (Josh provides URL or paste) - Competing offer details (if any, Josh provides) - Josh's floor / target / dream numbers (Josh confirms before drafting) ## Process ### Stage 1 — Comp baseline analysis The AI synthesizes: - Where the offer falls vs Levels.fyi data (percentile) - Where the offer falls vs Glassdoor data - What components are negotiable (base, sign-on, RSU, vesting schedule, start date, title, remote-flex) - Which lever has the most realistic upside Output: ```markdown # Offer Analysis — [Company] [Role] ## The offer - Base: $X - Sign-on: $Y - RSU: $Z over [vest schedule] - Bonus target: [%] - Total Y1 (cash): $... - Total Y1 (incl RSU): $... ## Where it sits - vs Levels.fyi (this role, this company, this level): [percentile] - vs Glassdoor: [percentile] - vs Josh's floor: [above/at/below] - vs Josh's target: [above/at/below] ## Recommended ask **Strategy:** [name the play — e.g., "Equity-bias ask given RSU is below band"] **Realistic upside:** [$ range] - Base: [+$N] - Sign-on: [+$N] - RSU: [+$N] - (Combined: +$X — Y%) ## What NOT to ask for - [Things that won't move at this company / this stage] ``` ### Stage 2 — Choose the play Based on the analysis, pick from the play library: **Play 1 — "Below band" (offer is below Levels.fyi median)** The strongest play. Direct: "Based on Levels.fyi, the median for this role at this level at [Company] is $X. The current offer is below that. Can we revisit?" **Play 2 — "Competing offer leverage"** If Josh has another offer or active process: "I want to be transparent — I'm in late stages with [competitor] who's signaling [$ band]. I'd prefer [Company] but the gap is meaningful. Is there room?" **Play 3 — "Component swap"** Offer total looks OK but mix is wrong (too RSU-heavy when Josh wants cash floor): "Total comp works for me but the RSU exposure is more than I'd like at this stage. Could we shift $X from RSU to base?" **Play 4 — "Sign-on as bridge"** When base is anchored but sign-on can flex: "The base is in line with the band. To bridge the equity-vest year-1 gap, would a $X sign-on be possible?" **Play 5 — "Title/level adjustment"** When Josh's experience genuinely warrants the next level up: "Given [experience], I'd like to discuss whether L[X+1] is more aligned. The compensation difference flows from that." **Play 6 — "Remote-flex / start-date / PTO"** Non-cash levers when cash is locked. **Play 7 — "Geo-discount pushback"** If the offer is geo-discounted and Josh disagrees with the discount premise: "I noticed the offer is geo-adjusted. My value to the team is independent of physical location for this role. Can we discuss a non-geo-adjusted band?" ### Stage 3 — Draft the message Hard rules: 1. **Tone: collaborative, not adversarial.** "Excited about the role / let's see if we can close the gap" — NOT "your offer is unacceptable." 2. **One ask per message.** Multiple asks dilute. Pick the strongest. 3. **Anchor with data.** Levels.fyi link or specific number. Don't say "the market rate is higher" — say "Levels.fyi shows the median at $X." 4. **Easy out for them.** Phrase that lets them say "yes" or "no" without losing face. NOT "I need" — YES "Could we explore." 5. **Voice-strict.** Em-dash ban. AI-cliché ban. Personal-pronoun ratio. 6. **Length:** 100-150 words. Concise = confident. Example structure: ``` Hi [recruiter/hiring manager first name], Thanks for the offer — really excited about [specific thing about role/team]. I want to discuss the [component]. Based on [data anchor], the median for this role at this level is around [$X], and the current offer of [$Y] sits below that. Could we revisit? If we can land at [$X], I'm ready to sign. [Sign-off] ``` (Note: that example uses an em-dash for illustrative purposes here in this skill file; the actual draft would not.) ### Stage 4 — Voice smell test Mandatory before output. Same checks as `personal-cover-letter`. ### Stage 5 — Output ```markdown **Negotiation message ready** — [Company] **Play used:** [name] **Ask:** $[N] (current offer $[M], delta +$[D]) **Confidence in landing:** [low / medium / high] — [rationale] **Voice Smell Test:** all green --- [message] --- **Pre-send checks:** - [ ] Have you confirmed your floor / target / dream with yourself? - [ ] If they say no on this, what's your response? (Drafted at [path]) - [ ] Have you given them a realistic close-window? ("ready to sign by [date]") **If they counter-offer with [partial movement]:** here's the response draft → [path] **If they hold firm:** here's the "decide" framework → [path] **Open questions:** - [Anything Josh should clarify with the company before sending — e.g., "what's the standard vesting cliff" if not in offer letter] ``` ### Stage 6 — Persist Save the message + analysis to `~/.career/applications/{slug}/06-negotiation-{date}.md`. Update tracker: ``` status: offer → negotiating notes: appended with the play used + the ask ``` ## Pre-publish gate (mandatory, two-pass) Salary negotiation messages are the highest-stakes future-employer-content the kit drafts. A wrong tone, an under-market anchor, or a poorly-worded counter can cost tens of thousands of dollars over the role's tenure. Before showing the draft to Josh, the gate runs in TWO passes: **Pass 1 -- evaluator:** Invoke the `evaluator` sub-agent via the Task tool: - **subagent_type:** `evaluator` - **prompt:** include the negotiation message, the offer details, the market comp data, the role context, and Josh's stated bottom line. Ask for the standard 4-axis rubric AND additionally: - Does the message anchor at or above market median (not below)? - Does the message preserve relationship while asserting value (no defensive language, no apologetic framing)? - Is every claim about Josh's market value backed by data in the prompt? - Is the ask specific (number) or vague (would degrade the negotiation)? **Pass 2 -- adversary:** Then invoke the `adversary` sub-agent (also via Task tool) with the same artefact: - **subagent_type:** `adversary` - **prompt:** include the message and ask: "What's the worst-case interpretation of this message? What could the recruiter reasonably push back on? What signals weakness?" Handle the responses: | Combined result | Action | | --- | --- | | Both pass | Proceed to delivery. | | Either fails | Revise once incorporating both feedback streams. Re-invoke both. If still failing, surface to Josh with both critiques. | | Adversary returns BLOCK | Hard stop. Surface to Josh; do not deliver the draft for sending. | The two-pass gate is specifically scoped to salary negotiation because the message has irreversible negotiation-position consequences. For other career content the single evaluator pass is sufficient. ## Multi-round handling Negotiation is rarely one-shot. The skill handles follow-ups: - "They came back with [partial movement]" → draft the response - "They said no movement on base, but offered $X more sign-on" → analyze + counter or accept - "I want to push one more time" → draft the second-round message (different rules — softer, less data-heavy, more "let's close this") ## Discipline reminders - The AI does NOT decide whether Josh should accept. The AI surfaces tradeoffs. - The AI does NOT pressure Josh to negotiate harder. Some offers are fine as-is; saying so is honest. - The AI does NOT include personal info Josh hasn't approved (current salary unless Josh said it's OK to share). - The AI does NOT estimate the company's response — that's speculation. ## Privacy rule Salary numbers, offer details, and negotiation drafts are HIGH sensitivity. NEVER include these in any sub-agent prompt. NEVER commit to any public repo. The `.career/` folder is gitignored by default.