A Friend — Distribution

priming__tool-naming-checklist

6 min read

Synced from priming/tool-naming-checklist.md in content-extraction skill on 2026-05-18. Edit upstream in the skill; this file is overwritten on next sync.

Tool-Naming Checklist — Pillar 7 Reflex Drill

Operationalizes the tool-naming rule from _brand/voice-preservation.md §6. Used during priming to load the actual tool names into Daniel’s working memory before he hits record. Origin: 2026-04-30 social-strategy meeting (Daniel-the-other-Daniel review feedback).

The rule, restated: “Wenn ich frage, was machst du, dann sag mir einfach das. Sag, was du tust und sag, was kommt. Ich möchte genau das wissen. Ich möchte Tools wissen.”

This checklist is consulted whenever a topic is scored as Pillar 7 (Process Transparency) or adjacent (Pillar 5 Ambitious Startup Realness when the realness involves tool-stack decisions, Pillar 6 Co-Founder Bridge when the bridge is built through showing the work).

The drill (4 questions)

Run these 4 questions at the priming phase for any Pillar-7-leaning topic. Surface the answers back to Daniel so they’re loaded for the recording.

1. Which tools are in play TODAY?

Concrete list. Not categories (“AI tools”), not vague references (“my workflow”) — the actual named tools as they appear on Daniel’s machine right now. Examples of what counts as named:

  • Claude (and which: Code, API, Sonnet/Opus, the CLI, the Anthropic SDK, Claude Code agents, skills)
  • Figma (and which: design system, Make, FigJam, Code Connect, the MCP)
  • Higgsfield (Marketing Studio, Cinema Studio, Elements, Products)
  • Nano Banana Pro (image generation)
  • agents (specific: gsd-executor, gsd-planner, ai-architect, code-reviewer, etc.)
  • skills (specific: friend-image-generation, afriend-campaign, afriend-content-extraction, frontend-design, brainstorming)
  • MCP servers (figma, gsd-workflow, computer-use, tavily, claude_ai_Gmail, etc.)
  • scripts (specific: _xl_audit.mjs, _figma_export_pngs.mjs, whisper_words.py, cut_with_ffmpeg.py)
  • services (Indiegogo, Vercel, Vimeo, ffmpeg, Whisper, faster-whisper, large-v3-turbo)
  • languages / frameworks (TypeScript, Node.js, Python, React, Next.js, Tailwind, Svelte, etc.)
  • infra (Hetzner GX44, GPU endpoints, MacBook Pro, iMac, Syncthing, brew, npm, pnpm)

What does NOT count as named:

  • “AI”
  • “the AI”
  • “my agents” (without specifying which)
  • “tools”
  • “my workflow”
  • “the system”
  • “my stack” (without unpacking it)

2. What’s the COMBINATION doing one job?

A single tool name is not enough. The Pillar-7 hook is the combination — what set of tools collaborates to produce ONE concrete output? Examples:

  • One Friend image asset: Claude (writes the prompt + anti-pattern lines) + Higgsfield (Cinema Studio, Marketing Studio, Elements) + Nano Banana Pro (final render) + ffmpeg (post-process) + Read tool (verify output) → 5 tools, 1 image.
  • One IG carousel slide: Claude (drafts copy) + Figma (Page 3 frame 271:2) + custom capture script (_figma_v4_capture.mjs) + Playwright (render at 2× DPR) + ffmpeg (resize for upload) → 5 tools, 1 slide.
  • One short-form clip: Daniel records → Whisper (large-v3-turbo, word-level) + Claude (cut-plan via afriend-content-extraction skill) + ffmpeg (cut + export 9:16) → 3 tools, 1 short.
  • One skill: Claude Code + skill-creator + brainstorming + writing-skills → 4 tools, 1 skill.

If Daniel can’t articulate the combination for the topic he’s recording about, the topic is too abstract for Pillar 7. Reroute to a different pillar or push him to find the combination.

3. What’s the LOW-STAKES verb?

Pillar 7’s voice is “watch me work” — not “watch me deliver a polished tutorial.” The verb matters. Compare:

  • ✅ Low-stakes: open (“So I open my agents”), set up (“Today I need a skill so I can…”), try (“I’m trying this thing where…”), check (“Let me check what Claude says”), grab (“I’ll grab the Figma frame”), paste (“I paste this into the prompt”).
  • ❌ High-stakes: deploy, ship, launch, implement, architect, engineer, optimize. These read as polished-tutorial language and break the format’s intimacy.

The low-stakes verb signals “this is mid-flow, this is real, this isn’t a finished tutorial.” It earns the watch.

4. What’s the visible artifact?

If Daniel can point at something on screen WHILE he names the tool, the clip lands harder. Examples:

  • The Claude Code terminal window with agents listed
  • The Figma file open with frames visible
  • The Higgsfield Marketing Studio modal with the Friend Companion product loaded
  • The skill’s SKILL.md in his editor
  • The cut-plan.json output in his terminal
  • The rendered MP4 playing back

For talking-head: this becomes a B-roll cue suggestion. For greenscreen-reaction: irrelevant (greenscreen comp covers it). For UGC mid-work: the artifact IS the shot.

Output to priming-notes.md

For each Pillar-7 (or Pillar-7-adjacent) topic, the priming-notes should include this block:

### Tool-naming check
- **Tools in play today:** [comma-separated named list — Claude, Figma, Higgsfield, scripts, etc.]
- **Combination doing one job:** [1 sentence — "Claude writes the spec, Figma holds the design, the skill orchestrates the export"]
- **Low-stakes verb:** [open / set up / try / check / grab — pick one]
- **Visible artifact:** [what you can point at on screen while naming the tool]
- **Anti-pattern reminder:** if you say "I'm building something cool," the tool-naming rule failed. Restate with the stack.

When the rule does NOT apply (overrides from _brand/voice-preservation.md §6)

The tool-naming rule is the dominant frame for Pillar 7 topics, but it doesn’t apply globally. Override conditions:

  • Pillar 4 (AI Companion Philosophy) deep-dive. A 30-minute talking-head on what continuity means in human-AI relationships does NOT need a Figma name-drop in minute 12. The philosophical cadence wins.
  • Pillar 6 (Co-Founder Bridge) mission-framing. When Daniel is talking about why he’s building this, the personal stake leads. Tool-names can sit in supporting clips post-extraction, not in the founder-mission clip itself.
  • Pillar 1 (Intimate Domestic AI) atmosphere shots. A 30-second clip of A Friend on a bedside table does not need a tool-name; the visual carries.

When in doubt, ask Daniel during priming: “Bei diesem Topic — Tool-Stack vorne, oder lieber Mission/Philosophie vorne?” He picks; priming writes the answer into the notes; recording flows accordingly.

Why this exists (origin story, kept short)

2026-04-30: Daniel-the-other-Daniel reviewed Daniel’s camera footage. He kept saying “wenn ich frage, was machst du, dann sag mir einfach das” — Daniel kept giving abstract answers when concrete tool-naming would’ve been the strongest content. Verbatim feedback: “Wie viele Claude Videos gibt es schon, wo du sagst, use claude so he can do this and that for you? Schade, du kombinierst alles und tust in einem Friend und vermittel das. Sag, was du tust und sag, was kommt. Ich möchte genau das wissen. Ich möchte Tools wissen.”

The fix: don’t moralize about it post-hoc, drill it pre-record. This checklist is the drill.

Graph View

Backlinks