uinaf/ffss

Strip AI tells from writing, a diff, or a test suite: detect machine patterns (puffery, stock vocabulary, crutch punctuation, filler, narrating comments, defensive boilerplate, tautological or overmocked tests, repetitive setup), rewrite with a concrete human voice, and self-audit. Use when asked to unslop, de-slop, humanize, or clean up text, docs, commit messages, tests, or a change before it ships. Do not use for correctness review (slopguard's lane), fact-checking, or translating.

Deliver a completed, verified change as a change request on the repository's forge: branch discipline, conventional commits, push, open the change request from the repo's template, attach the clearest visual evidence, and return the URL. Use when asked to deliver, ship, file, or open a pull/merge/change request for finished work, or when a slopmachine run reaches DELIVER. Do not use to implement or review changes, merge or enable auto-merge, or record slopmachine evidence by itself.

Runs one independent review of completed local changes, pull requests, branch diffs, or commits through the installed slopguard Go CLI using Codex, Claude, Cursor, or Grok; gathers authoritative acceptance criteria, selects or honors one provider, validates findings, applies scoped fixes, verifies and reruns accepted fixes, and safely reports reproducible CLI defects. Use when the user asks for a slopguard, a review of their code, pull request, or changes, an automated PR or second-model code review, a final tool-backed review, or review closeout after builder verification. Do not use as builder verification or a multi-reviewer panel.

Runs a structured, deterministic implementation workflow via the installed slopmachine CLI: clarifies intake, gates human release, runs build and verify, records independent review evidence, and delivers a reviewed artifact. Use when the user says /slopmachine, run this plan, execute the task list, ship this in slices, implement with checkpoints, walk the plan end to end, build it with human gates, or do a governed multi-step implementation. Do not use for ad-hoc edits or planning-only work.

Babysit an open change request through review and CI to a settled outcome: observe checks and reviewer feedback, verify every bot claim before acting, fix real findings without scope creep, answer threads with commit hashes, and merge once green unless asked to hold. Use when asked to babysit, monitor, or watch a pull/merge/change request, or after slopcourier delivers one. Do not use to create change requests or run a review yourself.

Audit and improve repository guidance, lifecycle commands, and runner infrastructure for dependable autonomous work. Use when making a repo agent-ready, agent guidance lacks orientation or proof paths, agents cannot boot or verify, setup still needs a human, or a runner must finish tasks unsupervised. Do not use for reviewing an existing diff, an ordinary builder self-check in a healthy repo, or prose-only documentation cleanup.

Audit, compress, restructure, and update repo documentation and durable agent-facing artifacts such as AGENTS.md, README.md, docs/, specs, decisions, and runbooks. Use when changes risk doc drift or when docs need current-state cleanup or context-efficient agent-first rewriting. Do not use to create tactical work plans, epics, or tracker tickets, or for code review, runtime verification, or boot/readiness infrastructure setup.

Turn an agreed change, specification, or conversation into a durable, resumable work plan in the repository's preferred tracker. Use when asked to plan future work, write or make a plan, save a plan, create an epic or tickets, break work into tickets, decompose work for parallel execution, or prepare work for another session. Do not use for routine in-session checklists, implementation, general documentation cleanup, or open-ended discovery where requirements still need substantive product decisions.

Whip a rambling response back into terse labeled deltas: outcome first, counts and identifiers over adjectives, at most six lines, one next action. Use when the user says /wat, wat, too long, tl;dr this, stop novelizing, or wall of text, and from then on whenever a drafted reply starts narrating process instead of reporting deltas. Do not use for cleaning committed artifacts, docs, or diffs (slopclean's lane), or for shortening content the user asked to be exhaustive.