Slushpile

---
name: slushpile
description: 'An adversarial job search that gets better as it goes. Set up a job-search workspace, search a careers board and score roles against the realistic applicant pool, build a resume and cover letter, review them with seven agents that return a verdict per submission channel, then write what the review finds back into the profile. Invoke with /slushpile and name what you want to do.'
disable-model-invocation: true
license: MIT
---

<!-- Generated by scripts/sync_docs.py. Do not edit this file. -->

# Slushpile

A job application pipeline. This skill is a router: the pipeline itself lives in
ten skill files and eight agent definitions under `.slushpile/`.

Read the matching file in full and follow it. Do not work from this summary —
the details are the product, and every rule in those files carries a failure
mode that this page does not have room for.

| The user wants to | Read |
|---|---|
| Set up a workspace, first time | `.slushpile/skills/onboard/SKILL.md` |
| Search a company's careers board | `.slushpile/skills/job-board-search/SKILL.md` |
| Find a warm path into a company | `.slushpile/skills/outreach/SKILL.md` |
| Surface experience they never wrote down | `.slushpile/skills/explore-experience/SKILL.md` |
| Build a resume and cover letter for a role | `.slushpile/skills/application-builder/SKILL.md` |
| Review materials before submitting | `.slushpile/skills/adversarial-review/SKILL.md` |
| Strip AI tells from prose | `.slushpile/skills/removing-ai-tells/SKILL.md` |
| Redesign the resume and letter templates | `.slushpile/skills/redesign-templates/SKILL.md` |
| Report on the search and check its calibration | `.slushpile/skills/status/SKILL.md` |
| Ask how any of this works | `.slushpile/skills/help/SKILL.md` |

## Before anything else

Check for `preferences.yaml` in the working directory. If it is absent, the
workspace has not been set up, and every other skill here depends on it. Run
onboarding first.

## The review pipeline without subagents

Cursor has no subagent dispatch. The review runs its seven personas sequentially
in one context.

Read each definition from `.slushpile/agents/`, adopt it, write its report out
in full, then move to the next. The order is: triage screener, requirements
analyst, ATS simulator, fatigued reader, pool analyst, hiring manager,
contrarian.

The first five are supposed to be blind to each other. In a single context they
are not, and a specialist that has already read the triage verdict will drift
toward agreeing with it. Writing each report out completely before starting the
next is what limits the bleed. Do not skip that.

The hiring manager gets all five specialist reports. The contrarian gets
everything.

## The one rule that matters most

Nothing in the pipeline hardcodes a fact about the user. Every personal fact —
compensation floor, relocation, work authorization, clearance, targeting — lives
in `preferences.yaml`, and every factual claim about their career lives in
`profile.md`. Read those. Never assume a fact about the user that is not in one
of them, and never write one into an application that you cannot point to.

## What this never does

It does not submit anything. No skill here touches an application portal, an
email, or a form. It writes files that the user reads and sends themselves.