菜单
Adversarial critique of a plan, design, or decision — produce the 5 sharpest criticisms a hostile-smart opponent would raise. Complement to /steelman. Use before committing to a major decision or when you suspect you're too attached to a plan.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Detect when a task is being repeated often enough to be worth automating, then propose the right automation — a PROGRAM (deterministic script) or a SKILL (reusable judgment procedure). Analyzes the current session, the memory/ record, or an observation log; classifies program-vs-skill; estimates payoff; drafts the automation. Proposes, never auto-builds. Use when you catch yourself or the user repeating a multi-step task, or to mine an observation log periodically.
Make a Wick agent fully portable between machines — consolidate a host's auto-memory into the Wick memory/ layer, disable the shadow portably, and relativize all paths and tooling. Guided, gated, archive-don't-delete. Use when moving an agent to a new machine, or to converge a single-machine agent onto one governed memory layer. Full reference in PORTABILITY.md.
Reflective pass over memory/ files — flag duplicates, stale facts, conflicts, and consolidation candidates. Proposes edits, never applies without confirmation. Use monthly or when /audit flags drift.
Produce a one-paragraph narrative summary of a CHANGELOG.md version entry — suitable for release announcements, blog posts, or social copy. Preserves technical specifics (file paths, command names, counts) rather than vaguifying them. Defaults to the latest version; optionally takes a version like "v1.0.0". Refuses to invent content if the version isn't in the file.
Extract structured fields from a source — paper, web page, API doc, or UI screenshot — and save a queryable record to memory/catalog/. Domain-agnostic. Use to build a personal index of things you'll want to find again, compare side-by-side, or reference in later decisions. Complements wick-research (which finds) and wick-tldr (which summarizes) by adding a structured-index layer.
Force base-rate thinking before a probability estimate. Identify reference class, historical rate, and adjust from there — not from intuition. Use before /calibrate or whenever you catch yourself or someone else estimating a probability from vibes.
| name | wick-red-team |
| description | Adversarial critique of a plan, design, or decision — produce the 5 sharpest criticisms a hostile-smart opponent would raise. Complement to /steelman. Use before committing to a major decision or when you suspect you're too attached to a plan. |
| license | MIT |
If /steelman is "construct the strongest version of a position you disagree with," wick-red-team is the inverse: take a position you're committed to, find its weakest seams, attack them.
Adopt a hostile-but-smart perspective. For the given plan/design/decision:
Produce 5 critiques, ranked by severity. For each, offer a mitigation — or name the trade-off explicitly as an accepted risk.
## Critique 1 — [severity: critical / high / medium]
**Attack:** [the specific flaw]
**Why it bites:** [conditions under which it breaks]
**Mitigation:** [what would harden against it — or "accepted risk because X"]
## Critique 2 ...
(5 total)
## Net recommendation
Ship / Revise / Rethink — with reasoning
Premortem analysis, weaponized. Inversion (Munger) — "how would I guarantee failure?" Gate 4 (Adversarial Convergence) — if red-team and steelman both reach the same concern, that concern is real.
Blunt, not cruel. The goal is to stress-test the plan, not to moralize about the planner. Name-calling and strawmanning are both failures of the skill.
wick-red-team followed by /decide → stress-test before committing/steelman followed by wick-red-team → full adversarial examinationwick-red-team before a major presentation → surface objections the audience would raise