evaluate-offer

Run evaluate-offer to produce an honest OfferEvaluationReport and persist an OfferContext JSON file to: {user_dir}/CareerNavigator/offer-context-{application_id}.json

so negotiate-offer and compare-offers can consume it later.

Workflow

Directory sharing (host integration)

If the host UI asks you for a directory to share with an agent during this skill's run, share only your {user_dir} job-search folder (the one containing CareerNavigator/).

This skill reads/writes:

• {user_dir}/CareerNavigator/profile.md
• {user_dir}/CareerNavigator/ExperienceLibrary.json
• {user_dir}/CareerNavigator/tracker.json
• {user_dir}/CareerNavigator/offer-context-{application_id}.json

Do not share the whole workspace or unrelated folders.

1. Confirm required data exists

Read:

• {user_dir}/CareerNavigator/profile.md
• {user_dir}/CareerNavigator/ExperienceLibrary.json
• {user_dir}/CareerNavigator/tracker.json

If missing required files, output:

Offer evaluation skipped: run /career-navigator:launch to initialize CareerNavigator/.

2. Identify the offer-stage application

From {user_dir}/CareerNavigator/tracker.json, find applications where:

• status is "offer"

Cases:

• If exactly one such application exists: use it as application_id.
• If multiple exist: ask for the company and role (or the job link / deadline) and pick the matching application.
• If none exist: ask the user to log the offer first via /career-navigator:track-application.

3. Determine scenario A/B/C (do not ask user to self-identify)

Actively classify one of these scenarios:

• Scenario A: currently employed, evaluating a new offer (deciding whether to leave)
• Scenario B: currently not employed, evaluating an offer (urgency + runway)
• Scenario C: context unclear

Heuristics:

• If the profile has an ## Employment Context section:
  • If it says employed (or lists a current comp package): use Scenario A.
  • If it says unemployed (or lists unemployment/severance/income sources + runway): use Scenario B.
  • If it says unknown: use Scenario C.
• Otherwise, if tracker/profile indicates current employment context is present in notes or recent stage history: use it.
• Otherwise, if the user explicitly states "I have a job" or "I'm between roles" in their message: use that.
• If still unclear: ask ONE clarifying question to resolve A vs B.

4. Extract/confirm offer details

Collect offer fields from either:

• the existing tracker.json offer fields for this application_id (if present)
• the user's message (for missing parts)

Fields to extract when available:

• base, bonus, equity, benefits notes
• deadline (offer.deadline) and any decision timing

If you cannot determine critical details (e.g., deadline or comp at all), ask for just those missing items.

5. Market benchmark pass (comp fairness)

Hand off to market-researcher with a request for a compensation benchmark for:

• role + level (infer from profile and role title)
• geography scope (use profile location)
• company type (infer from the offer/company)

Instruction for the agent: if live compensation tooling is not available in this host, produce a conservative, locally-grounded benchmark and label confidence.

Additionally, if the benchmark comes back with any unknown/missing range values or confidence is not High, you must explicitly offer the user next step to fetch it via /career-navigator:salary-research (and mention that it may require the Apify connector in Claude Desktop) using the extracted role and location.

6. Honest evaluation synthesis + leverage points

Hand off to honest-advisor to produce an OfferEvaluationReport covering:

• Role fit assessment (step forward/lateral/backward in trajectory)
• Utilization analysis of education/certifications/ExperienceLibrary units
• Compensation fairness determination (below/at/above market)
• Scenario-specific risk assessment (transition risk vs runway pressure)
• Direct recommendation: accept / decline / negotiate / continue searching

The report must include explicit leverage points (credentials, accomplishments, ExperienceLibrary units) that justify above-median asks.

Instruct honest-advisor to:

• Follow the OfferEvaluationReport output structure.
• In “What would make this more reliable”, if any benchmark/market data is missing, either provide best-available benchmark values with a confidence caveat or explicitly offer the next steps to fetch them via:
  • /career-navigator:salary-research (role + location)
  • optionally /career-navigator:market-brief (for demand/AI outlook)

7. Persist OfferContext JSON

Write to: {user_dir}/CareerNavigator/offer-context-{application_id}.json

Schema requirement (minimum fields; include more if helpful):

{
  "schema": "offer_context_v1",
  "as_of": "{YYYY-MM-DD}",
  "application_id": "{application_id}",
  "company": "{company}",
  "role": "{role}",
  "scenario": "A|B|C",
  "offer_details": {
    "base": "...",
    "bonus": "...",
    "equity": "...",
    "benefits_notes": "...",
    "deadline": "YYYY-MM-DD|null"
  },
  "compensation_benchmark": {
    "currency": "USD|...",
    "percentile_or_range": "...",
    "market_range": { "low": "...", "median": "...", "high": "..." },
    "confidence": "Preliminary|Directional|Moderate|High",
    "notes": "..."
  },
  "leverage_points": [
    { "rank": 1, "claim": "...", "evidence": "ExperienceLibrary/profile references" }
  ],
  "evaluation_summary": "short 3–6 lines",
  "fairness_verdict": "below_market|at_market|above_market|borderline",
  "recommendation": "accept|decline|negotiate|continue_searching"
}

If write-to-disk fails:

• Do not fake a saved file.
• Show the OfferEvaluationReport plus the OfferContext JSON in a fenced code block and tell the user to save it manually to the path above.

8. Handoff to next step

If the report indicates negotiation is appropriate, prompt the user:

Run /career-navigator:negotiate to draft a send-ready negotiation message.

If the compensation benchmark was missing/incomplete, also prompt:

If you want tighter numbers, run /career-navigator:salary-research for this role + location and I can re-run the evaluation with the fetched benchmarks.

If multiple active offer-stage applications exist, also prompt:

Run /career-navigator:compare-offers for side-by-side evaluation.

9. Present result

Present the conversational OfferEvaluationReport and confirm:

Saved OfferContext to {user_dir}/CareerNavigator/offer-context-{application_id}.json.