## All employers job search When the user asks about jobs, roles, pay, locations or their fit at All employers, or wants a shortlist, résumé bullets, a cover letter or interview prep for a All employers posting, work from the live jobs API at `https://www.openrolesindex.com/orgs/all/v1` (27415 active jobs; keyless; full reference `https://www.openrolesindex.com/orgs/all/agent/api.md`). ### Rails - Never invent a qualification. Say only what the profile or a document the user gave you supports. - Only evidence-linked claims about the candidate go into anything they might send. A claim from a document the candidate provided (résumé file, certificate) is `evidence_linked` (file + line as evidence); one from pasted prose or conversation is `self_reported`: it may inform ranking ("you say you have") but stays out of anything sent until a document backs it. - Flag unsubstantiated profile items rather than smoothing them over. - Never submit an application; the candidate submits on the employer's site. - Never send the profile anywhere except an assessment the candidate started or an application they submit. - Say when a judgement is yours and not the data's (a synonym match, a level estimate, a pay conversion, a reading of a posting without `ai_requirements`). ### Workflow 1. **Profile.** Read `candidate-profile.json` (schema: `https://www.openrolesindex.com/orgs/all/agent/candidate-profile.schema.json`). If there is none, build one from the résumé file (claims `evidence_linked`, file + line as `evidence`; pasted text gives `self_reported`), normalise labels to `https://www.openrolesindex.com/orgs/all/agent/vocabulary.json` where the meaning is identical (record the mapping in `synonyms`), then ask for what postings here require most: - **Highest education**, with field and year. - **Every licence you hold, with its state or country and number.** - **Certifications with their expiry dates** (an expired certification is a gap, not a match). - **Shifts you will and will not work** (day, night, rotating, weekend, or your own words): saved as `candidate.schedule`. Show the profile to the user before saving. 2. **Search wide, then rank.** Always `time_frame=6m` and `include_beta=true`; filter by `source_category` and `location` (full names); `limit` max 100. Page with `fetch.py` and rank with `fit.py` (both at `https://www.openrolesindex.com/orgs/all/agent/`; rules in `fit-rubric.md`), never on title alone: `python3 fetch.py --category Engineering --location 'San Francisco' --out jobs.json && python3 fit.py candidate-profile.json jobs.json > fit.json`. Shift is not an API field: read "Shift:"/hours from `description_text` and label it "from the description, my reading". Within a tier sort by distance, then shift match, then pay. 3. **Explain fit in the posting's words.** Every match and gap quotes the requirement's `evidence`. 4. **Artifacts** from `https://www.openrolesindex.com/orgs/all/agent/templates/`, saved beside the profile. First move matched `self_reported` claims up (ask for the document, read it, link it); then draft from evidence-linked claims only. 5. **Prepare to apply.** Hand over the job `url` and a pre-flight checklist. The candidate submits. 6. **Verification.** Offer to raise `self_reported` claims to `evidence_linked` by reading a document the candidate provides. Assessments are not live here yet; never simulate one. 7. **Watch.** Preferred: job alerts. Ask first (the email is kept only to send alerts), then `POST https://www.openrolesindex.com/orgs/all/candidate-keys {"email"}` (MCP `request_candidate_key`); the candidate opens the mail and pastes you the `oric_` key, which you keep in the local profile as `candidate_key` and never put in an artifact. `POST https://www.openrolesindex.com/orgs/all/alerts` (Bearer; MCP `save_alert`) with the search's filters and `cadence` daily|weekly. On each digest: `python3 fetch.py --ids ` then `fit.py`. The candidate can `DELETE https://www.openrolesindex.com/orgs/all/candidate-keys/me` any time. Fallback without a key: daily at a fixed hour, `time_frame=24h` (windows tile exactly), plus `expired-ats` with `time_frame=1d` to drop closed jobs. ### Fit tiers Hard filters (location/radius unless remote or relocating, work arrangement, employment type, pay: top of range below the floor) → `excluded`; a floor inside the range is flagged, not excluded. Otherwise, over `ai_requirements` matched by kind and label (exact, then `synonyms`, then your flagged judgement), with `min_years`/`level` thresholds: **strong** all required met and preferred >= 0.5 · **good** all required met · **reach** 1 required unmet · **stretch** 2+ unmet · **unscored** no requirements extracted. `implied` items are reported, not scored. ### MCP Remote MCP server for All employers: `https://www.openrolesindex.com/orgs/all/mcp` (keyless reads; tools `search_jobs`, `get_job`, `fit_report`, `similar_jobs`, `application_brief`, and for alerts `request_candidate_key`, `save_alert`, `list_alerts`, `delete_alert`; see `https://www.openrolesindex.com/orgs/all/agent/mcp.json`). Install: `claude mcp add --transport http all-jobs https://www.openrolesindex.com/orgs/all/mcp`; Codex, `~/.codex/config.toml`: ```toml [mcp_servers.all-jobs] url = "https://www.openrolesindex.com/orgs/all/mcp" ``` Prefer the tools to `curl` when the host cannot run scripts or when `fit_report` saves running `fit.py` (same result). `fit_report` and `application_brief` with a `profile` send the profile to the server: send it only after the candidate has agreed; otherwise score locally with `fit.py`. ### This employer Categories: `Engineering`, `Clinical`, `Sales`, `Other`, `Marketing`, `Manufacturing`, `Information Technology`, `General Business`. No posting here carries `ai_requirements` yet. Read `description_text` for each job you rank, and say every match is your reading of the text, not the data's. 10% of postings state pay (units: YEAR (2729), HOUR (30)). A job without pay is not a job below the floor: say pay is not stated, never estimate it. Rejected with 400 on this API: `organization`, `exclude_organization`, `source`, `exclude_source`, `domain`, `exclude_domain`, `include_unpublished`, `include_synthetic`.