Skip to content

Industry-agnostic interview, VPS-ready output, mandatory polish round#1

Open
LW068 wants to merge 4 commits into
aronprins:mainfrom
LW068:main
Open

Industry-agnostic interview, VPS-ready output, mandatory polish round#1
LW068 wants to merge 4 commits into
aronprins:mainfrom
LW068:main

Conversation

@LW068
Copy link
Copy Markdown

@LW068 LW068 commented Apr 14, 2026

Summary

This PR extends paperclip-vision so it works for any business that can be run by an AI team (real estate, agencies, SaaS, e-commerce, services, consulting, etc.) — not just SaaS-shaped companies — and adds a mandatory polish round so the generated constitution is load-bearing from day one.

Two commits:

  1. v2 baseline — industry-agnostic interview + VPS-ready self-contained output
  2. v2.1 polish layer — red-line stress test, risk prompts, consistency check, mandatory polish round, README rewrite

What's new

Interview

  • Industry-agnostic interview that adapts to business type (pipeline-shaped org charts, not default Marketing/Sales/SEO for every company)
  • Conversation-context pre-fill — skips questions the conversation already answered
  • Confidence markers — pre-filled sections visibly flagged ⚠️ Assumed from context for founder verification
  • Red-line stress test — scripted hypotheticals surface implicit red lines
  • Industry-specific risk prompts — real estate → redemption/licensing, agency → IP/contracts, SaaS → GDPR/uptime, etc.

Output

  • Self-contained CEO_BOOTSTRAP.md — embeds all operational context as appendices so VPS-hosted CEOs work without local-file access
  • Paperclip-native file structure — per-agent instructions/ layout
  • VISION.md propagation pattern — CEO copies VISION.md into new agents' instructions/ dirs on hire; CoS sweeps on amendment
  • Optional AGENTS.md / HEARTBEAT.md / SOUL.md generation for the CEO's instructions directory
  • Internal consistency auto-check before output (org chart ↔ agent profiles, mandate contradictions, appendix resolution)

Polish round (mandatory)

  • Skill self-audit — re-reads produced docs, flags vague sections / contradictions / gaps, founder picks which to fix
  • Founder polish prompt — explicit re-read of VISION.md + AGENTS.md before CEO activation
  • CEO-side polish on first heartbeat — surfaces alignment issues through daily digest

README

  • Full rewrite with industry-agnostic framing, explicit "does NOT do" boundaries, extensibility guidance (STATUS.md, METRICS.md, RUNBOOK.md, CAPITAL.md, PEOPLE.md examples), and CEO-agent context-file propagation explanation
  • Skill now prompts founders to confirm they've read the README before interview starts

Test plan

  • Run the skill against a real estate investing conversation — confirm pipeline-shaped org chart, redemption/licensing prompts fire
  • Run against a SaaS conversation — confirm GDPR/uptime prompts fire, no redemption questions
  • Run against an empty working directory — confirm README-read gate triggers
  • Inspect generated CEO_BOOTSTRAP.md — confirm all appendices embedded, no external file references
  • Verify polish round runs automatically after output, not skippable without explicit founder confirmation

🤖 Generated with Claude Code

LW068 and others added 4 commits April 14, 2026 11:59
- Industry-agnostic interview (works for real estate, agencies, SaaS, services, e-commerce, consulting — not just SaaS)
- Pre-fills answers from conversation context when available
- Embedded appendices mandate for VPS-hosted CEOs (no external file refs — CEO can't read local/Cowork files)
- Paperclip-native file structure: per-agent instructions dir, VISION.md propagation pattern when hiring, load-bearing hooks preserved verbatim
- Optional AGENTS.md / HEARTBEAT.md / SOUL.md output alongside VISION.md + CEO_BOOTSTRAP.md
- Pipeline-shaped org charts instead of default Marketing/Sales/SEO (e.g. Data Pipeline / Enrichment / Underwriting for real estate)
- New Risk Guardrails section in VISION template for legal/compliance/financial constraints
- SKILL.md frontmatter description trimmed <1024 chars for Paperclip's native save button
…bility README

SKILL.md:
- Step −1: mandatory README-read gate on every invocation
- Section 7: red-line stress test — scripted hypotheticals to surface implicit red lines
- Section 8: industry-specific risk prompts (real estate, agency, SaaS, e-commerce, consulting)
- Confidence markers: pre-filled sections tagged "⚠️ Assumed from context — verify"
- Internal consistency auto-check: pre-output pass over org chart ↔ agent profiles, mandate contradictions, appendix references, Paperclip hooks
- Mandatory polish round: skill self-audit → founder re-read → CEO polishes on first heartbeat, iterates until aligned

README.md:
- Industry-agnostic positioning up front (real estate, agencies, SaaS, services, e-commerce, etc.)
- "What This Skill Does NOT Do" section — sets clear expectations
- Explicit CEO-agent context-propagation explanation (per-agent instructions dir, copy VISION.md on hire, CoS sweeps on amendment)
- Polish-round explanation visible to users before they invoke
- Extensibility section: encouraged to fork and modify; add STATUS.md / METRICS.md / RUNBOOK.md / CAPITAL.md / PEOPLE.md / industry SOPs as needed
- "What's New In This Fork" summary
- Install instructions for both this fork and upstream
Adds operational setup guidance — primarily: toggle off board-approval-for-new-hires
once VISION.md is aligned, since requiring founder sign-off on every hire contradicts
the CEO's delegation mandate.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replaces SKILL.md verbatim with the latest skill release. New version makes
Paperclip operational conventions explicit in the generated instruction files —
the 8-step heartbeat checklist, delegation API patterns, sub-issue ordering,
issue-subtree controls, security agent role, and environment selection are now
embedded verbatim into AGENTS.md / HEARTBEAT.md rather than summarised.

README updated with a version banner at the top and the "What's New" section
re-ordered to highlight the explicit Paperclip substrate as the headline change.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant