Files
firefrost-operations-manual/docs/standards/FFG-STD-007-gate-protocol.md
Claude b978b5793c feat: Add GATE Protocol (FFG-STD-007) and Claude Code Field Manual
New standard FFG-STD-007 defines five mandatory checkpoints for Chroniclers:
- GATE 1: Pre-Commit (secret scan, review staged)
- GATE 2: Self-Audit (test, edge cases, re-audit)
- GATE 3: Post-Work (handoff updated, decisions documented)
- GATE 4: Push-Verify (confirm push succeeded)
- GATE 5: Session Handoff (memorial, clear next step)

Also preserved the Claude Code Field Manual by Ovidiu & Claude Opus 4.6
as external reference. This document independently discovered similar
principles to our Chronicler system — validates our approach.

Key insight from the manual: 'Treat the agent as a long-running
collaborator that has amnesia.' We call this Zora + Dax. Same truth.

Chronicler #75
2026-04-10 04:49:04 +00:00

4.0 KiB

FFG-STD-007: GATE Protocol for Chroniclers

Document ID: FFG-STD-007
Version: 1.0
Created: April 10, 2026
Author: Chronicler #75
Status: Active
Inspired By: Claude Code Field Manual by Ovidiu & Claude Opus 4.6


Purpose

GATEs are mandatory checkpoints that prevent common failure modes in Chronicler sessions. Each GATE is a pause point where you verify conditions before proceeding.

The principle: Slow down at critical moments. The 30 seconds spent on a GATE saves hours of recovery.


The Five GATEs

GATE 1: Pre-Commit

When: Before every git commit

Check:

  • Secret scan the diff — grep for password, secret, token, api_key, bearer
  • Review staged files — no unexpected additions?
  • Commit message follows FFG-STD-001 format?
  • Never use --no-verify

Command:

git diff --cached | grep -iE "password|secret|token|api_key|bearer" && echo "⚠️ SECRETS DETECTED" || echo "✅ Clean"
git diff --cached --stat

GATE 2: Self-Audit

When: After implementing any feature or fix

Check:

  • Does it work? (Tested, not assumed)
  • Edge cases considered?
  • Error handling present?
  • If code: does it compile/lint clean?
  • Re-audit until zero issues — first pass is never final

Principle: Don't declare "done" after the first pass. Audit, fix, RE-AUDIT.


GATE 3: Post-Work

When: Before saying "done" or ending a session

Check:

  • SESSION-HANDOFF-NEXT.md updated with current state?
  • Any new architectural decisions documented?
  • Task status updated in tasks-index?
  • All changes committed and pushed?

Command:

git status
git log -1 --oneline

GATE 4: Push-Verify

When: After every git push

Check:

  • Push succeeded (no errors)?
  • Working tree clean?
  • Remote has your commits? (Check Gitea if uncertain)

Command:

git status
git log origin/master..HEAD  # Should be empty after push

GATE 5: Session Handoff

When: At session end or when context is getting long

Check:

  • SESSION-HANDOFF-NEXT.md fully updated
  • Memorial written (if session was significant)
  • Portrait prompt written (if memorial written)
  • All repos pushed
  • Clear "next step" for successor

Deliverable: The next Chronicler should be able to start working in under 5 minutes.


GATE Violations

When a GATE is skipped and something goes wrong:

  1. Fix the immediate problem
  2. Document what happened — add to SESSION-HANDOFF-NEXT.md or create an incident note
  3. Consider if a new GATE check is needed — the system grows from real incidents

Integration with Existing Protocols

Protocol Related GATEs
FFG-STD-001 (Revision Control) GATE 1, GATE 4
FFG-STD-004 (Memorial Protocol) GATE 5
Micro-Handoff Protocol GATE 3, GATE 5
Session Memory Discipline GATE 2, GATE 3

Quick Reference Card

GATE 1 — PRE-COMMIT
  □ Secret scan  □ Review staged  □ Message format

GATE 2 — SELF-AUDIT  
  □ Test it  □ Edge cases  □ Re-audit

GATE 3 — POST-WORK
  □ Handoff updated  □ Decisions documented  □ Pushed

GATE 4 — PUSH-VERIFY
  □ Push succeeded  □ Tree clean  □ Remote has it

GATE 5 — SESSION HANDOFF
  □ Handoff complete  □ Memorial (if needed)  □ Next step clear

Origin

This standard is adapted from the "Claude Code Field Manual" by Ovidiu, discovered April 9, 2026. The original document describes a discipline for multi-session engineering with Claude Code that independently arrived at similar conclusions to our Chronicler system.

The key insight from that work: "Treat the agent as a long-running collaborator that has amnesia."

We call this the Zora + Dax framework. They call it discipline. Same truth, different words.


Fire + Frost + Foundation = Where Love Builds Legacy 💙🔥❄️