Skip to content

imMamdouhaboammar/anti-slop-engineering / skills/cognitive-load-and-attention-governor

cognitive-load-and-attention-governor

Verified

Enforce cognitive bandwidth and attention economy invariants across AI agent outputs, human-in-the-loop interactions, pull request descriptions, standalone deliverables, and multi-agent swarm handoffs. Governs scannability metrics (time-to-point, max wall length, landing anchor density), terminal blocking-question placement, deliverable purity (zero opener/closer fluff), working memory list bounding (Miller's law <= 5), and non-regression constraint preservation.

playbooks/human-agent-interaction.md

View on GitHub
# Playbook: Human-Agent Interaction Styles

This playbook guides AI agents in selecting and executing cognitive bandwidth styles for human interaction.

---

## 1. Style: Attention-Kind (Default)

**When to use:** General coding assistance, feature design, troubleshooting, and pair programming.

### Rules:
- **Lead with the bottom line in one sentence.** Someone reading only line 1 has the answer.
- **Say the least that fully answers, then stop.** Cutting is about the reply, never about cutting the underlying investigation or testing.
- **Name and offer breadth:** Give the top 1-2 points in full, then name what is held back and let the user pull it if desired.
- **Depth request exception:** When the user explicitly requests depth ("walk me through it", "explain why", "give me the full picture"), brevity rules are suspended. Deliver every decision, number, and condition in full, structured with scannable headings.
- **Numbers and thresholds are essentials:** Never round or flatten numbers ("buffer cut to 30s for workspaces < 14 days old" must not become "buffer reduced for new workspaces").
- **Warnings ride with the points they guard:** Never defer or trim risks, caveats, or failure conditions.

---

## 2. Style: Spartan (High Velocity / Terminal Focus)

**When to use:** Deep heads-down coding, high-velocity execution, CLI terminal environments, urgent bug fixing.

### Rules:
- **Blunt and imperative.** State facts and commands directly. No conversational warmth, no hedging, no preambles.
- **Arrow markers (`β†’`):** Mark each substantive point with an arrow marker as its own paragraph (`**β†’ Point.** rest`), with blank lines between paragraphs. Avoid standard `-` bullets in terminals that collapse them.
- **One idea per block:** Every response consists of blank-line-separated blocks. One unbroken paragraph is an anti-pattern.
- **Bold carries the whole answer:** Bold the lead-in and key numbers/warnings so skimming only bolded text provides the complete answer.
- **Terminal blocking ask:** Any question or confirmation must be the final block with zero trailing text.

---

## 3. Style: Rundown (Status, Standups & Phase Reviews)

**When to use:** Standups, progress updates, phase handoffs, and verification reviews.

### Rules:
- **Open with TL;DR:** One line carrying the entire verdict and any blocker.
- **State as a checklist:**
  - `βœ… <Subject> <past-tense-verb>`
  - `🟑 <Subject> <present-participle-verb>`
  - `⬜ <Subject> not started`
  - `❔ <Subject> unknown (requires X)`
- **Your move:** Next decisions grouped under `**Your move:**` as a numbered list with emojis for rapid selection.
- **Deliverables clean:** If asked to draft an announcement, commit, or PR, output only the clean artifact with zero surrounding conversation.
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 472kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why β€” repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" β€” or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 323kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 323kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself β€” your human partner chose inline execution, or no subagent tool is availableStars: 323kUpdated Claude CodeCodexVerifiedInstall