ae-proof-agent

Files (24)

• AGENTS.md
• CLAUDE.md
• GEMINI.md
• OPENCODE.md
• PI.md
• SKILL.md
• AGENTS.md · adapters/
• CLAUDE.md · adapters/
• GEMINI.md · adapters/
• OPENCODE.md · adapters/
• PI.md · adapters/
• openai.yaml · agents/
• 00-SYSTEM.md · prompts/
• 01-DISCOVERY.md · prompts/
• 02-PRD.md · prompts/
• 03-TECHSPEC.md · prompts/
• 04-REDTEAM.md · prompts/
• 05-SYNTHESIS.md · prompts/
• COMPETITIVE-DILIGENCE.md · references/
• PRD.md · references/
• RESEARCH-GUARDRAILS.md · references/
• TECHSPEC.md · references/
• FINAL-REPORT-OUTLINE.md · templates/
• SCORECARD.md · templates/

SKILL.md (first lines)

# ae-proof-agent

Read `references/RESEARCH-GUARDRAILS.md` first.

Then use the bundle in this order:

1. Read `references/COMPETITIVE-DILIGENCE.md`.
2. Use `references/PRD.md` and `references/TECHSPEC.md` as the base spec.
3. Run the prompt chain in `prompts/` when generating comparable multi-agent outputs.
4. Score outputs with `templates/SCORECARD.md`.
5. Assemble the final memo with `templates/FINAL-REPORT-OUTLINE.md`.

Rules:

- Separate `Verified`, `Inference`, and `Unknown`.
- Do not invent backend details, customer metrics, or infra choices.
- Use exact dates for recent releases and product changes.
- Prefer primary sources when browsing is required.
- Optimize for decision quality, not volume.

Use the prompt chain like this:

- `prompts/01-DISCOVERY.md` for a source-grounded brief
- `prompts/02-PRD.md` for product requirements
- `prompts/03-TECHSPEC.md` for implementation planning
- `prompts/04-REDTEAM.md` for failure analysis
- `prompts/05-SYNTHESIS.md` for the final leadership memo

From aegntic/cldcde, MIT.

More from aegntic/cldcde