agentcrew
claudeFour everyday skills: polite (tone rewriting), humanizer (removes AI-sounding patterns), explain (plain-language re-explain), first-principles (decision decomposition).
7 skills
checkpoint: one-step session close, with a full fact check via /checkpoint full (neat-freak merged in), an automatic lite mode for small sessions, and a lessons step that turns your corrections into rules. Assumes you keep a daily log and your notes or projects in git; if you don't, skip this one.
Plugin agentcrew-closure from marketplace agentcrew-harness in agentcrew-academy/harness-starter-kit. Commands read from the repo's manifest — today.
checkpointClose a session once: daily log, fact reconciliation (master files, project docs, memory), selective commit/push, and remote verification. Use for /checkpoint, end of session, save, push, commit and push. Full: full reconciliation, knowledge sync, memory cleanup. Lite mode: inline, no subagent, for
explainRe-explain the previous reply or a specified topic in plain language a high-schooler could follow, no jargon. Triggers: /explain, plain English, I don't get it, ELI5, too much jargon.
first-principlesFirst-principles decomposition: ask whether the requirement should exist, strip to ground truths, rebuild the minimal solution. Triggers: first principles, first principles rethink, reason from the ground up, why should this be done this way, blind spots, unknowns inventory.
humanizerRemoves AI writing patterns(去 AI 味)from English or Chinese text without changing what it says. Use before sending outward drafts. Triggers: 去 AI 味, AI 痕跡, 太像 AI, humanize, de-slop.
phantom-pushbackUse when a reply or document carries caveats that attribute a belief, assumption, expectation, or worry to the reader ('you may be assuming...', 'one honest caveat...', 'this plan treats X as...') and each one needs checking against what was actually said or written — typically a closing paragraph t
politeThree-mode tone rewriting: A Warm & Empathetic (customer service replies, delivering bad news, apologies, reassurance, declining a request; too cold, too blunt, soften it up); B Formal Business (outward-facing emails to an unspecified audience, announcements; too casual, too abrasive, more formal, b
review-loopRuns a multi-round human review loop over a long AI-written document (spec, plan, report, proposal) without drift between versions. Use whenever the user is reviewing a draft you produced and wants changes, says things like next version/v2/v3, I recorded my feedback/here's the transcript, revise thi
Four everyday skills: polite (tone rewriting), humanizer (removes AI-sounding patterns), explain (plain-language re-explain), first-principles (decision decomposition).
7 skills
Asks you to confirm before any tool sends an email, chat message or post, so nothing goes out unseen.
Blocks the end of a turn when the AI claims 'verified' or 'not found' without having run a check or a search, or when it edited files after the last check.
Blocks file writes that contain emoji and tells the AI what to use instead.
Silently records which files each session edited, so checkpoint stages only this session's changes when several sessions share a repo.
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.